Friday, May 17, 2024

Environment and Sustainability: New Possibility of Growth in Switzerland - Juniper Publishers

 Environmental Sciences & Natural Resources - Juniper Publishers


Abstract

Switzerland's tourism industry, while economically successful, faces challenges in environmental impact and visitor experience authenticity: this paper explores the potential of rural and agricultural tourism as a new avenue for sustainable and environmentally friendly growth. The idea is to propose a framework for developing low-impact, high-engagement experiences that connect tourists with nature, local culture, and agricultural practices. This approach fosters environmental responsibility, economic diversification in rural areas, and the creation of unique value for tourists seeking a deeper connection with Switzerland.

Keywords: Environmental impact; Sustainable tourism; Rural tourism; Nature; Agrotourism; Switzerland

Introduction

In an historical period where populations and civilizations are moving more and more into residential agglomerations, urban centers and metropolises the desire to "escape" from everyday life and the need of being reunited with the nature become a necessity, increasingly felt by a growing number of individuals [1]. At the same time, some consumers give priority to elements linked to low environmental impact, to sustainability, for respecting of nature and organic products rather than only the price or the easy availability of goods [2].

From these increasingly trends, new possibilities linked to agriculture and rural sectors are born with the research of a total immersion “in the green”. Alongside the institutional vacations that we all know, we can see some structures that allow - or that require - participation in agricultural life: in addition to a traditional approach related to holiday farms, horseback riding or bird watching we see new realities in which tourists can have an active role in the organization of rural life, being responsible for looking after farm animals or taking care of a vegetable garden or participating in the harvest [3].

But what do we mean when we talk about "agricultural tourism" and "rural tourism"? Without wishing to conduct a linguistic analysis in this study of the two terms, we could define them for convenience as the various forms of tourism directly connected to territorial resources and which find their main component in rural culture. Therefore, it is not just a matter of tourism towards rural areas, but of an original approach to low-impact tourism, which includes a fully use of a territory [4]. This certainly means allowing agriculture, in all its forms, to become the protagonist of a holiday from the recovery and enhancement of traditions to the consumption of typical products, from the visit of cultural and artistic heritages to active and experiential participation in rural life and activities. All by retracing ancestral values such as circularity, sharing and sustainability [5].

The Territory

Switzerland's tourism industry has long been a cornerstone of its economy, contributing significantly to GDP and employment [6]. However, the traditional model of mass tourism often comes at the expense of environmental degradation and cultural commodification [7]. In response to growing concerns about overtourism and climate change, there is a growing consensus that a more sustainable approach to tourism is needed. Switzerland, with its commitment to environmental conservation and high-quality tourism experiences, is ideally positioned to lead this transition towards sustainability [6].

Switzerland Tourism has defined 13 visitor types, according to their needs, with the purpose of having a better segmentation of the tourist demand and one of those is the so-called “Nature Lovers”, searching for gentle and authentic interaction with nature as a way of recharging their batteries [8]. Switzerland, with its central position in Europe and its diverse geography, has many natural resources to offer to visitors seeking sustainable tourism, who are attentive to the environmental and social impact of tourism activities but always looking for authentic and responsible experiences. In addition to that, many of the traditional tourism activities are already inherently sustainable, taking place outdoors and allowing you to appreciate the natural beauty of the region - the Confederation has one of the highest percentages of renewable energy in the world and is working towards climate neutrality by 2050 [9].

Methodology

This research adopts both descriptive as well as analytical approach for carrying out the research results. The will is to investigates the innovative approaches to sustainable tourism development in Switzerland analysing real cases of the territory to understand what is being done and what the future developments of the sector could be. Several field research, interviews and insights have been conducted not only for the realization of this poster but for a more structured series of articles based on how effectively agriculture and rural tourism can help in the development of a more sustainable tourism.

Results

Switzerland has a long tradition linked to organic agriculture and sustainable food products, such as cheese, wine, and chocolate and, according to our research, these are the elements to be exploited to create a new value proposition involving sustainable tourism [10]. The cheese is one of Switzerland's signature products and having the possibility of making this culinary delicacy in a dairy can be a unique and funny experience for tourists [11]. For example, several dairies open their doors to tourists, allowing them to see how cheese is made and to actively participate in the production process. During the visit, tourists can also actively participate in the production of products, getting hands-on and helping the dairies prepare the cheese. This can be a very entertaining, engaging and educational experience, as visitors can see the cheesemaking process up close and have a first-hand experience of the art and tradition of cheese making [12]. Do we consider the satisfaction of eating a homemade cheese made by yourselves? (Figure 1)

Another favourite practice of tourists traveling to Switzerland is sleeping on straw: this is a unique authentic experience in which tourists can try a form of traditional accommodation that may not be available elsewhere. This kind of ancestral experience allows visitors to immerse themselves in the local culture and learn about the traditions and practices of the area they are visiting. "Sleep in Straw" is a Swiss association which, among various activities, aims to promote this unique experience. On their website there are listed - at the date of publication of this article - 59 farms that offer overnight stays in straw beds (Figure 2).

The approach of various wine producers in the Mendrisio area - located in the southern part of Switzerland, in the canton of Ticino, near the border with Italy - is interesting, allowing tourists to have a unique experience, in close contact with nature, actively participating in the harvest. Several wineries welcome visitors and let them experience firsthand the magical festive atmosphere that traditionally accompanies harvest time by spending a day among the rows, picking bunches of ripe grapes, learning closely the first stages of winemaking. Enhancing the wine tourism experience leads to an enhancement in consumers' attitudes towards wine, their assessment of both extrinsic and intrinsic attributes, and their loyalty towards various wines. Moreover, segmenting consumers based on their level of wine tourism experience can assist wine marketers in comprehending their target audience and market direction [13].

Conclusion

Because of its history, its geography and its tradition, Switzerland has many opportunities to offer sustainable tourism thanks to its natural resources but, above all, thanks to a culture of sustainability that also passes through its agriculture and its sustainable food products. Switzerland can continue to be an ideal tourist destination for those seeking an authentic and sustainable experience but has to innovate new tourist proposition to be more captivating and contemporary like Regenerative Tourism (that aims to regenerate and restore the natural and cultural systems of local communities) or New Technologies (changing the tourism industry worldwide, including agriculture and rural tourism) with mobile apps and artificial intelligence that are leading to greater transparency, user engagement and personalized services for travellers.

By embracing this new approach, Switzerland can ensure the long-term viability of its tourism industry while preserving its pristine environment and unique cultural identity.

To Know More About International Journal of Environmental Sciences & Natural Resources Please click on:

For more Open Access Journals in Juniper Publishers please click on: 

Wednesday, May 15, 2024

ASPEN Study Case: Real Time in Situ Tomato Detection and Localization for Yield Estimation - Juniper Publishers

 Agricultural Research & Technology - Juniper Publishers


Abstract

As the human population continues to grow, our food production system is challenge. With tomato as the main fruit produced indoors, the selection of varieties adapted to specific conditions and higher yields is an imperative task if we are to meet the growing food demand. To assist growers and researchers in the task of phenotyping, we present a study case of the Agroscope phenotyping tool (ASPEN) in tomato. We show that when using the ASPEN pipeline, it is possible to obtain real-time in situ yield estimation without a previous calibration. To discuss our results, we analyse the two main steps of the pipeline in a desktop computer: object detection and tracking, and yield prediction. Thanks to the use of YOLOv5, we obtain a mean average precision for all categories of 0.85, which together with the best multiple object tracking (MOT) tested allow obtaining a correlation value of 0.97 compared to the real number of tomatoes harvested and a correlation of 0.91 when considering the yield thanks to the use of a SLAM algorithm. In addition, the ASPEN pipeline demonstrated to be able of predicting subsequent harvests. Our results demonstrate in situ and real-time size and quality estimation per fruit, which could be beneficial for multiple users. To increase the accessibility and use of new technologies, we make publicly available the necessary hardware material and software to reproduce this pipeline, which includes a dataset of more than 820 relabelled images for the tomato object detection task and the trained weights

Keywords: Tomato; Yield; Food production system; Phenotyping; Agricultural industry; Fruit detection

Abbreviations: MOT: Multiple Object Tracking; UGV: Unmanned Ground Vehicles; NN: Neutral Networks; LIDAR: Light Detection and Ranging; SFM: Structure from Motion; SLAM: Simultaneous Localisation and Mapping; ASPEN: Agroscope Phenotyping Tool; CNN: Convolutional Neural Network; GUI: Graphical User Interface; GFLOPS: Giga Floating-Point Operations Per Second; ROI: Region of Interest; RSE: Residual Standard Error

Introduction

As we approach the estimated inflection point of the world population growth curve UN, increasing global food availability is more important than ever. Especially with the current climate crisis threatening our food system Owino et al. [1]. Several strategies have been applied throughout the food production chain to address this issue FAO [2]. To this end, new methods have been tested across the agricultural industry to speed up results and increase efficiency, particularly in new technologies in a so-called fourth agricultural revolution, even though the impact of these new technologies is not clear Barret & Rose [3]. The vast majority of these new methods require large amounts of information obtained directly from the field or plants, in descriptive processes called phenotyping. Phenotyping is the activity of describing, recording or analysing the specific characteristics of a plant and due to the nature of this process and the required frequency, it is a time-consuming task Xiao et al. [4]. While the principle of phenotyping is not new, the quantity and quality of information that is today been generated has never been seen before, making imperative to reduce the time required for this task. Remote sensing has been used and its automation has already been demonstrated thanks to new algorithms and technologies Chawade et al. [5], even further opening up new opportunities for real-time data utilisation, what could save resources and further improve the industry as a whole Bronson & Knezevic [6].

There are several ways to automate phenotyping, with each path depending on the allocated budget and working conditions. For example, Araus et al. [7] divide these paths according to the distance to the target. Satellites can be used with fast data acquisition per m2, but with a trade-off between resolution and investment costs. Other methods closer to the plants, such as stationary platforms, allow for higher spatial resolution data, but are less flexible and more expensive to implement. A proven solution is the use of drones, which are more flexible than fixed platforms, but still not as flexible as they cannot work in covered crops. On the other hand, handheld sensors, manned or unmanned ground vehicles (UGVs) are more flexible than the aforementioned platforms and have a higher resolution, a lower initial investment, but can cover smaller areas than the previously mentioned methods. The effectiveness of this last category has been well demonstrated, especially in the fruit detection and localisation tasks e.g. Scalisi et al. [8], but affordable open-source alternatives are scare. The phenotyping subtask of fruit detection in images was initially based on shape and colour, until the advent of neutral networks (NN), which were particularly advanced after 2012 e.g. Hinton et al. [9]. These have led to more robust results in fruit detection.

Today, well-established algorithms such as RCNN Girshick et al. [10], Mask-RCNN He et al. [11] and YOLO Redmon et al. [10] are constantly used for research purposes and in production environments. For example, Mu et al. [12] show that when using RCNN, they could achieve a mean average precision at 0.5 intersection over union (IoU) (mAP@0.5) of 87.63%, which later correlates with the real number of tomatoes per image at 87%. Other authors, Afonso et al. [13]; Seo et al. [14]; Zu et al. [15] showed that when using Mask-RCNN, focused on the task of instance segmentation, they obtain a similar or higher average precision than Mu et al. [12], with values of mAP up to 98%, 88.6% and 92.84% respectively in each study. These previous works demonstrate the ability of the presented algorithms not only to detect objects, but especially to detect individual tomatoes in situ. A notable point of these works is that the comparability of their results is technically incorrect since each detection algorithm was trained on different image datasets. To compensate for this, a standardised dataset must be used and although some few datasets are freely available online to train machine learning algorithms in the task of tomato object detection, these are rarely used. Remarkable datasets are "laboro tomato" Laboroai [16] and "tomatOD" Tsironis et al. [17] due to its quality and availability.

Although the previously mentioned NN based algorithms have good detection rates, they are not capable of running in real time (more than 30 frames per second, FPS). For example, using a variant of R-CNN, Faster R-CNN, Seo et al. [14] achieve up to 5.5 FPS using a desktop computer equipped with a graphics processing unit (GPU) card (NVIDIA GTX 2080 ti), without mentioning the input size of the model. Thanks to the introduction of YOLO, Liu et al. [18] have shown that near real-time analysis is possible. In their case, the authors improved the YOLOv3 model by using a denser architecture and round boundary boxes that better fit the shape of tomatoes. These changes allowed them to achieve an F1 score, a weighted average of precision and recall, of 93.91% at a speed of 54 ms (18 FPS), compared to 91.24% at 45 ms (22 FPS) and 92.89% at 231 ms (4.3 FPS) for the original YOLOv3 and Faster R-CNN, respectively. In their case, images of 416x416 pixels were processed on a desktop computer equipped with a GPU (NVIDIA GTX 1070Ti). With a faster, more robust and more recent version of YOLO, YOLOv5, Egi et al. [19] achieve an F1 score of 0.74 for red tomatoes, which correlates at 85% with a manual count. Although no speed was documented in their work, the various algorithms of YOLOv5 are capable of running in real time at resolutions below 1280 pixels, depending on the system used (CPU vs. GPU) and model implementation Jocher et al. [20]. In addition, Egi et al. [19] demonstrate that the use of a state-of-the-art multiple object tracking (MOT) algorithm allows each individual object to be tracked along a video sequence.

For the tracking task, several MOT algorithms have been proposed, among which we highlight SORT Bewley et al. [21], bytetrack Zhang et al. [22] and OCSORT Cao et al. [23] as their code is publicly available, they have a high performance and they can run in real time in a common CPU unit even in the presence of multiple objects. Once an object has been detected, it needs to be located in space, which can be done in a number of ways. One simple way is to use RGBD cameras that contain a deep channel (D). This information can be used to distinguish the foreground from the background objects, which has been well demonstrated in tomatoes by Afonso et al. [13], allowing for object localization within frame, but missing the global position of the detected objects. Using an alternative methodology, Underwood et al. [24] show that it is possible to reconstruct a non-structural environment using Light Detection and Ranging (LiDAR) technology for within frame, together with GPS data for global localization in a post-processing method. Thanks to their method, they were able to locate and estimate almond yield at tree level with an R2 of 0.71.

The use of 3D reconstruction techniques has been less explored in greenhouses. Masuda et al. [25], show that tomato point clouds obtained from structure from motion (SfM) can be further analysed to obtain per plant parameters such as leaf area, vapour length using a 3D neutral network, Pointnet++ Qi et al. [26], with an R2 of 0.76 between the ground truth area and the corresponding number of points.

In a more advanced analysis, Rapado et al. (2022) show that by using a 3D multi-object tracking algorithm, that really in an RGB camera and LiDAR, they achieve a maximum error of 5.08% when localising and counting tomatoes at a speed of 10 FPS. Similarly, other authors have documented that by the year 2022, pipelines based on existing 3D neural networks are slower than 2D methods that really in additional sensors to obtain deep information (e.g. Afonso et al. [13], Ge et al. [27]. In addition, actual 3D neural networks have major limitations such as maximum input size, large number of parameters that make them slower to train, high memory consumption, and moreover, they are particularly limited by the lack of datasets for training reasons Qi et al. [28].Nevertheless, new low-cost 3D pipelines and datasets are constantly being released to increase the availability of this technology (e.g. Schunck et al. 2022, Wang et al. [29].

In order to correlate these detections with real yields, 3D localisation is required, and remarkably, simultaneous localisation and mapping (SLAM) algorithms have not been widely used in agricultural environments, possibly due to their lack of robustness Cadena et al. [30]. Previous SLAM methods are suitable for more structured environments, with clear corners and planes that allow incoming LiDAR scans to be aligned, which can be used for localisation (LiDAR odometry, LIO). A possible solution for unstructured environments is to use images for localisation (visual odometry, VIO), but this tends to fail in fast motion. Sensor fusion, a technique that fuses multiple sensors together, can provide more robust systems that can, for example, align incoming LiDAR scans when using VIO for navigation. This technology is better suited to environments that lack clear features, such as outdoor environments. Notable examples of these algorithms due its robustness and open source code include VINS-FUSION Qin et al. [31], CamVox Zhu et al. [32], R3Live Lin & Zhang [33], and FAST-LIVO Zheng et al. [34].

The low use of these technologies in the agricultural sector, either separately or together, could be attributed to several reasons, including the maturity of the technologies, budgetary reasons, and a knowledge gap between farmers and computer science e.g. Kasemi et al. [35]. The Agroscope Phenotyping Tool (ASPEN) aims to break the digital phenotyping barrier among agricultural researchers, thanks to a proven and affordable pipeline that can work in situ and in real time for fruit detection, allowing non-experts to use the tool. In this paper, we demonstrate that this pipeline: 1) allows 3D reconstruction of a non-structured environment using a SLAM algorithm, and thanks to this 2) can localise and describe tomato fruits in a traditional greenhouse thanks to the addition of an object detection algorithm. Most importantly, in order to increase the accessibility and use of this pipeline, we are making the necessary hardware and software to reproduce it publicly available, which we hope will help to bridge the gap between agricultural and computer scientists.

Materials and Methods

Hardware and Software

An ASPEN unit was used to evaluate the ASPEN pipeline (Figure 1). Although it is not the aim of this paper to discuss the configuration or selection of the equipment used, a brief description is given below. For more details, the reader is invited to refer to the online project repository (https://github.com/camilochiang/aspen, Chiang et al. in preparation). The ASPEN pipeline considers a set of input sensors connected to an embedded computer using the robot operating system (ROS, Stanford Artificial Intelligence Laboratory et al. 2018), version melodic in a gnome-based version of Ubuntu 18.04.6 LTS, with the aim of reconstructing and locating plants, fruits or diseases in situ in real time, where we here focus in the fruit case. ASPEN uses a specific selection of sensors and electronic components that may already be present in an agricultural research facility. To achieve this goal, the system relies on two main workflows, tightly coupled and orchestrated by an embedded computer equipped with a GPU (Jetson Xavier NX 16 Gb): the camera workflow and the SLAM workflow.

For the camera workflow, a synchronised RGBD without timestamp synchronisation (Realsense, R415 - 1920x1080 pixels at 30 FPS with synchronised depth) is processed with a convolutional neural network (CNN) object detection technique based on the RGB. Once that a model has been selected and object detection per frame has been performed, each object is identified and tracked using a multi-object tracking (MOT) algorithm were an unique ID is assigned. Finally, once the detected tracked object passes a region of interest (ROI) of the field of view and it is confirmed as a unique object who have not been register before, its localisation within the image is transferred to the 2D to 3D estimation node. Using the localisation given by the MOT algorithm for each object, the 2D to 3D estimation node uses the deep (D) frame information to estimate the dimensions (mm) of the tracked object and its localization with respect the camera position. For each object detection, the minimum distance to the camera is extracted and then the actual diameter is calculated and used by a dimensional model (Figure 1) to convert to weight (g).

In addition to this workflow, two other sensors, an Inertial Measurement Unit (IMU, BMI088 bosh) and a Light Detection and Ranging (LiDAR, Livox mid-70, configured into single return mode) unit, as well as the RGB channels of the RGBD camera, are used in the parallel SLAM workflow who allow to locate each RGBD frame within a global mapping and therefore each tracked object in a 3D map. These sensors were choose due its low cost compared with similar sensors, and in case of the LiDAR especially due the extreme low minimum detection range (5 cm). The aim of this workflow is to reconstruct the environment in which the tomatoes are located and to provide a relative position for each tomato (with respect to the initial scanning point), which will then allow the detected tomatoes to be correlated with the handmade measurements. For this task, R3Live Lin & Zhang [33] was chosen as the SLAM algorithm, as it attaches new incoming points from the LiDAR unit (10 Hz) using the IMU (200 Hz) and image information and does not really only use LiDAR features for this task and can run in real time (faster than 30 FPS). These characteristics, shared with other similar visual odometry algorithms (VIO), show in our preliminary research to work better in agricultural environments in collaboration with SLAM algorithms that rely only in LiDAR odometry (LIO) (data not shown), potentially due to the clear lack of features (corners, planes) in a so-called "unstructured environment", which makes LIO algorithms more difficult to converge. To allow reproducibility, the input from all sensors are recorded within the ASPEN unit. A simple graphical user interface (GUI) is available to facilitate this task.

Experiments

To evaluate the ASPEN pipeline in the specific task of tomato detection and localisation, we started by training YOLOv5. Five different models (n, s, m, l and x) from the YOLOv5 family were trained at two different resolutions: 512 (batch size 20) and 1024 (batch size 6) pixels up to 300 epochs. These models differ mainly in the complexity of the model architecture, with the simpler models aiming to operate under resource-constrained conditions, such as mobile phones and embedded computers. This network was trained using 646 images for training and 176 images for validation, coming from our own datasets and other open source datasets (laboro-tomato and tomatoD). Regardless of the origin of the dataset, tomatoes were re-labelled in three different categories: immature, turning and mature tomatoes, with approximately 3500, 1000 and 900 instances of each category, respectively. After training, one of the resolutions and one of the models were selected for a posteriori use. For details of the dataset, the reader is invited to visit the online repository.

Once the model that met our requirements and had the best performance had been selected, two commercial-type greenhouses in the facilities of Agroscope (Conthey, Switzerland), with tomatoes of the Foundation variety grafted on DRO141, were scanned with an ASPEN unit on three consecutive harvest days in the middle of the production period of 2022. Each scan lasted a maximum of 12 minutes and was performed close to midday to ensure similar light conditions. Each greenhouse of approximately 360 m2 contained eight rows of tomato plants, each row 25 m long. The six central rows were scanned sequentially, with both sides of each row scanned before moving on to the next row. These rows were also divided into 3 blocks for other experiments, with buffer plants at the beginning, between blocks and at the end of each row. The scans were recorded as bag files using ROS. The resulting bag files were then transferred to a desktop computer (Lenovo ThinkPad P15, Intel core i9, GPU NVIDIA Quadro RTX 5000 Max-Q, 16VGb) for reproductive and posterior analysis. The analysis was automated with the aim of detecting tomatoes per block. To do this, the videos were first reviewed and pre-registered with timestamps of the transition between blocks.

To validate our results, after each scan we harvest all the tomatoes ready for marketing. Harvesting was done per bunch, usually from 4 to 5 tomatoes, which occasionally led to the harvesting of turning tomatoes. Harvesting took place either on the same day or the following day after each scan. To increase the spatial resolution of the validation data, each row was harvested side-by-side and each row was further divided into three different blocks, resulting in 216 validation points. This was incorporated into the analysis using a distance filter with the D-frame of the RGBD camera, ignoring any objects detected more than 50 cm from the ASPEN unit, as these correspond to elements in the background or on the other side of the row. For each harvest, the total weight was measured, including the weight of the pedicel. Differently, the number of fruits was counted per block only, regardless of the side of the row, giving 108 validation points. To build the size-to-weight model shown in Figure 2, after each scan, 100 tomatoes were harvested from the non-scanned rows belonging to the three categories mentioned above. These tomatoes were measured and weighted, and the model used later for yield estimation is shown in Figure 3.

To complement the validation of the ASPEN pipeline, three MOT algorithms were tested under similar implementation frameworks and parameters (Python 3.8): SORT Bewley et al. [21], Bytrack Zhang et al. [22], and OCSORT Cao et al. [23].The quality of the yield estimation results depends not only on good object detection, but also on correct tracking along the frames until each object reaches a region of interest (ROI), where it is counted. Independently of the MOT algorithm used, an estimated position, size and weight was calculated for each tomato detected. An example of the detection and reconstruction process is shown in Figure 4. The correlation of the three different MOT algorithms with weight and count in relation to the real harvest is shown in Figure 5.

Statistics

A priori and posteriori statistical analyses were performed using Python 3.8 Van Rossum & Drake [36] and the Statsmodels package (version 0.13.5, Seabold & Perktold [37]. A quadratic equation was fitted to the size-weight relationship (Figure 3), as this statistically fit the data better than a simpler relationship (data not shown). To estimate the correlation between crop yields, either in number or weight, a linear correlation without intercept was fitted between the manually measured data and the estimated data from the ASPEN pipeline, considering each crop subsample as a data point (n = 108 for the number task and n = 216 for the weight task). To evaluate the ability of the ASPEN pipeline to predict future yields based on previous measurements, we correlate the estimated number of tomatoes in the turning category with the following 3 harvests for each subsample as a data point (n = 108). Finally, to evaluate the task of size measurement, an f test of the size distribution was carried out within each category (Figure 6).

Results

Object detection

Within the ASPEN pipeline, the first task in the camera workflow is object detection (Figure 1), which requires a previously trained object detection model. As shown in Figure 2, when evaluating the task on the desktop computer using the family of models of the YOLOv5 algorithm (n, s, m, l and x models with 4, 16, 48, 109 and 207 Giga floating-point operations per second, GFLOPS), at a resolution of 512 pixels (px), an increase in the complexity of the model used allows a higher mean average precision at interception over union of 0.5 (mAP@0.5), which is particularly the case between the first two models (nano;n vs. small;s). Subsequently, more complex models (medium; m, large; l and extra-large; x) did not contribute to a higher mAP@0.5. In contrast to the lower resolution results, a higher resolution of 1024 px results in higher mAP@0.5 values for simpler models. At 512 px, the improvement in mAP values due to higher complexity was close to 2% between the two simpler models (n vs s), while a higher resolution contributed up to 5% improvement in mAP@0.5 values between the two n models.

Selecting the simplest model, YOLOv5n, also reduced the inference time from 8 ms to 25 ms compared to the more complicated model (x). Figure 2B shows the precision-recall curve of the selected model (YOLOv5s at 1024 px). The F1 values, a weighted average of precision and recall ranging from 0 to 1, were 0.941, 0.777 and 0.838 for the immature, turning and mature categories at mAP@0.5, with an average F1 value of 0.852 across categories. Irrespective of the category, the main difficulty was with precision measurement, suggesting a high number of false positives. Although the mature category had a similar number of cases to the turning category (around 900 compared to 1000), it is interesting to note that the turning category is still the most difficult to discriminate. On the other hand, the green category has a higher F1 value with more than 3500 instances.

Size to weight model and localisation

The next step was to investigate weight estimation using manual diameter measurements. For this purpose, a linear model represented by a parabolic function was used, as this one fitted our data better than other functions (data not shown). This correlation, with an R2 of 0.886, holds regardless of the ripeness of the tomatoes (data not shown) and when considering the production of layers of either small or large size, as shown in Figure 3. The average weight of the tomato samples was 147 ± 2 g (standard error, SE), which corresponded to the average weight of the harvested tomatoes during the scanning process.

Localisation estimation

To illustrate the localisation process, an example scan is shown in Figure 4. Figure 4A shows the object detection where different tomatoes are marked in boxes. These objects were then tracked using one of three different multi-object tracking (MOT) algorithms and once they passed a region of interest (blue line in Figure 4, they were registered, localised and measured in 3D space as shown in Figure 4B using the D channel from the RGBD camera, the size-to-weight model (Figure 3) and 3RLive. The selection of the region of interest (ROI) boundary was based on previous work in fruit detection (e.g. Borja and Ahamed, 2021) and an observed better object detection even in the presence of occlusions, as the objects were closer to the camera.

ASPEN pipeline validation

The number of tomatoes detected and their respective calculated weight is shown in Figure 5, in relation to the number of tomatoes harvested and their weight. It can be seen that both MOT algorithms of the SORT family underestimated the number and/or the total weight of tomatoes, while the bytetrack algorithm strongly overestimated both parameters. In addition, the bytetrack algorithm produced a significantly higher residual standard error (RSE) for both measurements compared to the SORT family algorithms. No statistical difference was found between the SORT algorithms independent of the measured variable. Independently of this, OCSORT was chosen as the best MOT algorithm due to a lower RSE. The size distribution of a manual measurement compared to the automated procedure is shown in Figure 6 for the OCSORT MOT algorithm. The distribution of measurements from the automated method did not differ from the manual method, regardless of the tomato category. On average, the ASPEN measurements were slightly lower than the manual measurements, but similar dynamics could be observed, with green tomatoes having higher mean diameter values (60 vs 56 mm) and a wider distribution, turning tomatoes having a lower mean value (62 vs 67 mm) and a skewer distribution, and ripe tomatoes also having lower mean values (64 vs 69 mm) and a similar distribution compared to the manual measurements. When correlating the number of turning tomatoes with the actual harvest and the next three harvests, the highest correlation was found when using OCSORT. Regardless of the MOT algorithm used, these correlations were weaker over time and have an increasing RSE. The third harvest was an exception, where a slight increase in the average correlation was observed (Table 1).

Discussion

The results presented here validate the use of ASPEN for tomato yield estimation. Although several previous studies have demonstrated the capability of image analysis using machine learning approaches, it was not until the introduction of YOLOv5 Jocher et al. [38] that real-time image analysis was possible. Mu et al. [12] showed that using R-CNN could achieve a mAP@0.5 of 87.83% when training on a category of tomatoes, and the detections correlated at 87% when compared to manual counting on the same images. Seo et al. [14] found 88.6% of tomatoes in images using a faster version of R-CNN: Faster R-CNN. In their case, they were also able to classify into six different categories, which took a total of 180 ms (5.5 FPS) per image on a computer equipped with a GPU. After the introduction of YOLOv3, near real-time results have already been achieved. Liu et al. [18] show that modifying YOLOv3 for the tomato object detection task allowed them to increase the F1 score from 0.91 to 0.93 with a small increase in inference time from 30 (33 FPS) to 54 ms (18 FPS) for images of 416 x 416 pixels. In their case, these changes were due to a denser mesh and a circular bounding box that allowed higher mAP@0.5. Using RC-YOLOv4, a more recent and modified version of YOLO, Zheng et al. [34] achieve an F1 score of 0.89 with a speed of 10.71 FPS on images of 416 x 416 pixels in a GPU equipped computer, suggesting that the improvement between YOLOv3 and YOLOv4 is mainly due to the gain in detection quality and not to the speed of the algorithm.

More recently, and similar to our work, Egi et al. [19] demonstrated that a flying drone with side view, using the latest YOLOv5 together with DeepSORT as MOT tracker, could achieve an accuracy of 97% in the fruit counting task in an average of two tomato categories, and a 50% accuracy in the flower counting task. Notably, their paper does not mention the speed of the various steps involved. These previous works demonstrate the capacities of previous and current algorithms for tomato fruit detection, where our work aligns with these results at similar F1 scores and shows how these capacities have increased over time and can be applied to the task of tomato fruit detection. Although not perfect, see Figure 4 for a clear tomato occlusion, we were able to correlate the number of tomatoes with the actual harvest to 97% in real time using YOLOv5 without prior calibration of the method, and thanks to the speed of the algorithm we were able to further improve the results. A limitation of YOLOv5 is the lack of subcategories, which could improve the detection efficiency. Training the same dataset with the same model and resolution (YOLOv5s), but with only one category, achieved a higher F1 score of 0.95 (data not shown) compared to three categories (F1 value of 0.852).

This suggests that our pipeline could be further improved by adding a second step classifier after the object detection algorithm, without losing real-time capacity. To further improve not only the count but also the weight correlation, it is also possible to use instance segmentation algorithms e.g. Zu et al. [15], Fawzia & Mineno [39], Minagawa & Kim [40]. This change may increase the accuracy of the weight model, as only the area of each tomato is detected, which should remove many errors in size measurement, especially those due to occlusion or overlap. So far, the speed of this task has been the limiting factor for real-time instance segmentation, but newer and faster algorithms may allow better results in our pipeline Jocher [20]. A negative effect of introducing an instance segmentation algorithm would be to increase the mathematical complexity of the size determination task, as it may be possible to fit a sphere into the D-frame Gené-Mola [41].

Several methods have been tested to determine the size and position of each fruit. Mu et al. [12] showed, similarly to our work, that it is possible to obtain dimensional features in tomatoes using an RGB camera, but due to the lack of a third dimension, their data was only displayed as pixels. Thanks to the addition of a deep (D) channel, Afonso et al. [13] were able to filter foreground objects from their Mask RCNN detections, while our work shows that we can not only filter foreground objects, but also obtain object characteristics in real time (Figures 4-6), which can be useful to study the growth dynamics of tomato fruits. In terms of speed, the use of the D channel to obtain sizes has been demonstrated to be the fastest method available in 2022. For example, Ge et al. [42] using the 2D boundary box output of an object detection algorithm together with the corresponding depth frame took between 0.2 and 8.4 ms compared to 151.9 to 325.2 ms when using a 3D clustering method. Similarly, Rapado et al. (2022) were able to reconstruct tomato plants using an RGB camera and LiDAR with multi-view perception and 3D multi-object tracking, achieving a counting error of less than 5.6% at a maximum speed of 10 Hz.

While other high quality methods have been tested in tomato plant reconstruction e.g. Masuda [25], these can be up to 100 times more expensive than lower cost and resolution methods Wang et al. [29] and cannot run in real time. Several studies have been carried out using SfM to evaluate lower cost 3D reconstructions in greenhouses, but thanks to the recent introduction of cheaper solid-state LiDAR technology, our pipeline is able to run in real time at a similar economic cost to SfM. The benefits of 3D reconstruction have been well demonstrated in tomato, e.g. Masuda [25] were able to correlate the actual leaf area and stem length of tomato plants with their respective number of points, which can be useful in the task of phenotyping. When using LiDAR technology, the chosen SLAM technique plays a crucial role. In our case, R3Live successfully reconstructed the unstructured environment on a desktop computer in real time (average of 24 ms for visual and LiDAR odometry), but it is important to mention that the algorithm has more than 25 parameters to be tuned and that under stress conditions (fast movements, camera occlusions and turning points) this one constantly fails to converge, weakening the whole pipeline. The main reason for the failure was identified as the lack of clear features, planes and corners, which are usually absent in unstructured environments, and further research is required e.g. Cao et al. [43]; Zheng et al. [34], especially when porting the pipeline to the embedded computer.

The robustness of the MOT algorithm and the selection of a good ROI are crucial for the object localisation task. In our case, with the same settings, both SORT algorithms perform better than Bytetrack, mainly due to a multiple ID assignment, demonstrating the importance of a good MOT algorithm selection for the yield estimation task. Although newer tracking algorithms have been tested in the tomato counting task e.g. Egi et al. [19], they can be slower than the simpler algorithms presented here, especially when tracking multiple objects. Regarding a good choice of ROI, Borja & Ahamed [44] show in pears that a ROI located in the central part of the image gives the best results in their case. In our case, we observe that a ROI located at 75% of the image field of view gives the best results, since objects are closer to the camera, allowing the detection algorithm to make better predictions and reduce the probability of occlusions. Regarding the SORT algorithms, both were able to predict the amount or weight of the crop per experimental unit (Figure 5), but in an underestimated way. This could be partly explained by technical reasons or more practical ones. On the technical side, the lack of detection due to occlusion (Figure 4) or fruit leaving the field of view before entering the ROI could contribute to the error.

Meanwhile, practical reasons include the fact that tomatoes were harvested by bunch, which includes the occasional turning of tomatoes and the weight of the pedicel (with an average value of 50 gr per bunch). Independently, the addition of the D channel proved to be useful in capturing the size differences between categories (Figure 6) and reduced the uncertainty of the weight model by about 1 kg for the SORT models when compared to the product of the uncertainty of the count model and the average tomato weight. Although no difference was found between the size distributions, the slight difference between the sizes of the categories shown in Figure 6 may have contributed to the uncertainty of the weight model, but further investigation is required as the sample sizes were extremely different (300 manually measured vs. 27000 digitally measured tomatoes). Finally, our pipeline demonstrates the ability to additionally localise and predict future harvest based on the turning category, which, similar to our previous correlation results, has a higher correlation when using the SORT family of algorithms. Further research is needed to validate these claims.

To our knowledge, the results presented are the first example of real-time detection, characterisation and localisation of tomato fruit in situ and without calibration. Several experiments have been shown to work in post-processing with other fruits e.g., Underwood et al. [24], and as a result, commercial platforms are already available e.g., Scalisi et al. [8], Ge et al. [42]. These platforms can perform similar work, but they generally require a site/crop pre-calibration and do not have the flexibility presented here. The advantage of pre-calibration is that images can be captured at a faster rate, linked to GPS coordinates and therefore faster scanning speeds could be achieved, resulting in a lower price per m3 scanned. Although this is an excellent approach for commercial orchards where GPS connectivity is available and decisions can be made a posteriori, real-time data acquisition and processing allows decisions to be made in real time and in the field. The open source pipeline presented adds the flexibility of a terrestrial laser scanner that can work not only outdoors but also indoors. In addition, the lateral view of the crop and the higher image resolution may allow early disease detection when the ASPEN pipeline is coupled with a multispectral camera [45].

Conclusion

The present study demonstrates the capabilities of the ASPEN pipeline in the detection, characterisation and localisation of tomato fruits. Thanks to a series of sensors, we were able to reconstruct the scanned environment in real time, opening the doors to new developments and possibilities not only for the task of fruit detection, but also for other real time visual related measurements (e.g. disease and pest detection). In this study, the ASPEN pipeline correlated with the actual number and weight of harvested tomatoes at 0.97 and 0.91, respectively, and although the pipeline is not perfect, possibilities for improvement were discussed, especially with the aim of reducing the uncertainty of the method. Thanks to the 3D reconstruction of the environment, other physiological measurements could also be automated (e.g. leaf area, plant volume), but further research is needed, especially to compare these results of an affordable 3D scanner with high quality scanners. We hope that the presented results will stimulate agricultural researchers to work with new technologies, and to inspire this, we make publicly available the hardware material and software necessary to reproduce this pipeline, which includes a dataset of more than 850 relabelled images and models for the task of tomato detection.


To Know more about Agricultural Research & Technology: Open Access Journal

To Know more about our Juniper Publishers

Thursday, May 9, 2024

Implants in Orthopedy - Juniper Publishers

 Orthopedic & Orthoplastic Surgery - Juniper Publishers


Abstract

Orthopedics is a branch of surgery that deals with disorders and conditions that involve the musculoskeletal system. Orthopedic surgeons use surgical and non-surgical agents to treat musculoskeletal trauma, spinal diseases, sports injuries, degenerative diseases, infections, tumors, and congenital disorders.

Keywords: Orthopedics; Bone; Ligament; 3D; CT

Introduction

The quality of human life can be dramatically improved with the use of biomaterials [1]. Rapidly advancing technologies are allowing new and improved biomaterials to be developed with unprecedented performance behaviour.

Biomedical Engineering

There has been significant improvement in technologies to reconstruct musculoskeletal defects as a result of trauma or disease [2]. During the last few decades, there has been widespread use of bone-banked, processed skeletal allografts to reconstruct large deficits of bone and cartilage with outcomes at intermediate follow-up providing 85% satisfactory results. However, there still is a significant incidence of nonunions and graft failures, which usually require additional surgical intervention and result in additional morbidity. Additionally, the cost and availability of graft materials and some immunological issues still have not been completely resolved. Although great strides have been made to improve materials and surgical techniques, the failure rate in these younger patients still approaches 10% in long-term follow-ups. The ultimate goal of any treatment that addresses musculoskeletal tissue loss is the restoration of the morphology and function of the lost tissue. The recent emergence of a new discipline, defined as tissue engineering, combines aspects of cell biology, engineering, materials science, and surgery with the outcome goal to regenerate functional skeletal tissues as opposed to replacing them. Repair and regeneration of skeletal tissues are fundamentally different processes. In many situations, scar, which is the result of rapid repair, can function satisfactorily, such as in the early phases of bone restoration. By contrast, regeneration is a relatively slow process that ultimately results in a duplication of the tissue that has been lost. Regeneration is rarely seen in adults but is evident in very young children. Such regeneration appears to recapitulate some of the key steps that occur in embryonic development. Our approach to musculoskeletal tissue regeneration is to use principles of tissue engineering that are based upon the premise that there are important constituents that distinguish the fetal environment from that in adults and by mimicking aspects of these fetal microenvironments, we can engineer the restoration of adult tissue. The basic component of any tissue engineering strategy is the use, either in combination or separately, of cells, biomatrices or scaffolds/delivery vehicles, and signaling molecules that provide the biological cues for the progression of cellular differentiation and its site-specific functional modulation. Significant issues remain for each component that must be addressed to develop successful and realistic tissue engineering treatment strategies. Central to our strategies is the need for cells. Significant issues that remain include the source of these cells, the number and density, and, most important, their age, phenotypic character, and developmental potency. We have put forth the hypothesis that mesenchymal stem or progenitor cells possess the appropriate developmental potential, are responsive to local cueing, and are capable of ultimately differentiating into the appropriate required phenotype. By contrast, adult differentiated cells are generally less responsive to mechanical and biological cues and may not be available in the appropriate quantities to achieve the desired tissue density.

Tooth

Tooth is a biological organ originating from ectomesenchymal cells composed of enamel, dentin, and viable pulp tissue which is altogether called as tooth organ [3]. These tissues usually arise from the interaction of oral epithelium and mesenchyme of cranial neural crest.

Bone

As a highly specialized and dynamic tissue, bone is characterized by its mineralized matrix, rigidity and hardness with certain degree of elasticity [4]. Bone provides support and protection to internal organs and also aids in locomotion.

Ligament

Ligaments are specialized connective tissues whose biomechanical properties allow them to adapt to and carry out the complex functions required of the body [5]. While ligaments were once thought to be inert, it is now recognized that they are in fact responsive to many local and systemic factors which influence their performance within the organism. Injury to a ligament results in a drastic change in its structure and physiology and may resolve by the formation of scar tissue, which is biologically and biomechanically inferior to the ligament it replaces. T1 weighted images are particularly useful at demonstrating the normal anatomy [6]. Ligaments will appear black against the adjacent fat which will be white. In case of injury, T2 weighted images will show edema in the soft tissues and if fat suppression is used then this can easily be differentiated from fatty structures. Therefore, T2 weighted images with fat suppression, or perhaps more sensitively, Fast STIR images should be employed. Because the anatomy of the lateral complex is variable, the choice of imaging planes is difficult. True axial images are particularly useful for looking at both the anterior and posterior tibiofibular ligaments. The anterior talofibular ligament will also be seen on most axial images although arguably an oblique axial running along the plane of this ligament may be more precise. Much more difficult is the calcaneofibular ligament. This is unfortunate as it is the most structurally important and therefore where we would like to image most accurately. The calcaneofibular ligament runs in oblique plane from the calcaneus running anteriorly and superiorly to the fibula. The angle varies with individuals and the shape of the hind foot. It is very difficult to judge the inclination of the best imaging plane to produce a true axial of this ligament. It is common that axial images will show the ligament on multiple slices, and it is difficult to follow its integrity even with MIP reconstructions. Alternative strategies are to place the foot in an equinus position, which elevates the calcaneus, making the calcaneofibular ligament a more horizontal structure. In this position a true axial is more likely to show the calcaneofibular ligament in its full length, but this may be a difficult position for the patient to achieve and hold, particularly if the ankle is painful. Therefore, it may be easier to examine the foot in a neutral position and incline the imaging plane with the anterior margin more cranial. The difficulty is how to assess the degree of angulation that would be required for an individual. Careful palpation of the ankle and judgement of the imaging plane by the examining technician or radiographer may assist. True 3D volume imaging of this region has an advantage that reconstructions can be made in different planes. However, 3D volume is most effectively achieved using gradient echo imaging and the contrast between the ligament and the adjacent structure is not as effective as it is on spin echo imaging. Therefore, 3D volume images are more difficult to interpret.

Reconstruction

A mechanically stable and bioactive substance would dramatically change the practice of reconstructive fields, such as orthopedic, plastic and oromaxillofacial surgery [7]. Percutaneous procedures with injectable, bioactive and resorbable cements could replace invasive treatments of acute fractures, chronic nonunions, and critical-sized bone defects. Management of soft tissue defects that also require mechanical strength, such as rotator cuff patches, anterior cruciate ligament (ACL) reconstruction, and cartilage or meniscal repair could likewise be performed with minimally invasive procedures and incur little functional loss during recovery.

For this reason, there has been considerable research in nanotechnology, which considers the biomaterial properties such as chemistry, charge, wettability, and surface roughness. These determine the extracellular protein interactions and mediate cell interactions at the tissue/matrix interface, which are critical for biocompatibility and longevity of the implant. In vitro research of surface morphology has suggested the importance of nanometer roughness. Up to four times the calcium-mineral deposition occurs when osteoblasts were cultured for 28 days in the presence of ceramics with grain sizes below 100 nm, compared with conventional alumina surfaces. Even greater osteoblast performance has been reported in grain sizes below 60 nm. This has been correlated to osteoblast interactions with vitronectin, which shares a linear dimension of approximately 60 nm. Multiple techniques are now being explored, such as e-beam lithography, polymer demising, chemical etching, cast-mold techniques and spin casting to fine-tune surface characteristic for optimal biologic interactions. In addition, three-dimensional (3D) printers can construct 3D organic-inorganic composite matrices with a defined internal architecture. These have also demonstrated osteoblast ingrowth and proliferation in vivo.

3D

Obviously, the use of the computer and associated software has benefited the orthopedic surgeons in other aspects, such as preoperative planning, preoperative 3D imaging, intraoperative computer navigation in total joint and spine surgery, besides trauma surgery, more recently, virtual intraoperative impingement and stability testing in ACL reconstruction in the field of sports medicine [8].

CT

Computed tomography, or CT, greatly facilitates 3D viewing of the internal morphology of soft tissue and skeletal structures [8].

Clinical Pathways

The adoption of clinical pathways in patient care has grown from the necessity of providing consistently high quality of care for an increasing demand for clinical services [9]. Clinical pathways are structured multidisciplinary care plans that detail the essential steps in the care of patients with specific clinical problems. Clinical pathways provide hospitals with a consistent template for patient care by creating a predetermined standardized approach to care that should be adhered to by each member of the healthcare team. Clinical pathways are especially suited to the high volume and elective nature of much of orthopedic surgery. In our specialty quality and efficiency must be optimized. To help achieve this clinical pathways are used as standard protocols. Each process, in a clinical pathway, is followed in order to ensure that the desired end results are achieved. The pathway also ensures that each patient is receiving optimum levels of care pre, intra-, and postoperatively. Clinical pathways are evidence based using the common international experience but must be adapted to the culture of any given hospital. Clinical pathways are effective because they standardize care, help develop measures for prevention of patient discomfort and harm and provide ongoing performance measures that promote effective and useful change in practice.

Adoption of clinical pathways can be met with skepticism and resistance from any member of the multidisciplinary team involved in patient care. Because clinical pathways standardize care they reduce reliance on individual decision making or traditional approaches to care. Every effort must be made in adopting new clinical pathways to educate and inform the multidisciplinary team of the evidentiary basis on which the principles of the new pathway are based. Physician, nursing, and administrative champions must work together to develop and institute new pathways. The process should be communicated in a completely transparent manner. The thought processes involved should be clearly documented, and every member of the patient care team must be trained and oriented to the new process. Following implementation documentation of important clinical indicators should be monitored and regular reports of outcome must be communicated back to the hospital staff. The pace of implementation must be geared to the tolerance of the staff at each individual hospital. Often implementation should be conservative with realistic expectations. As success is garnered more progressive modifications to the pathway based on real outcomes can be pursued. It is critical that the clinician champions involved in this process be sensitive and realistic as well as willing to devote their time and energy to the process.

Education

At a time of groundbreaking medical advances in the diagnosis and treatment of arthritis and musculoskeletal diseases, patient education has become an essential component in providing comprehensive care and in achieving positive clinical outcomes [10]. These advances, coupled with novel education delivery systems such as the Internet, have created consumer demand for information from patients, their families, and the general public.

Conclusion

Traumatological implants are used for the surgical treatment of fractures, deformities, and tumor diseases of the bones. In addition to products intended for the fixation of long bone fractures, trauma implants for the shoulder, hand, pelvis and hip are also produced. It should certainly be noted that innovations in orthopedics and traumatology serve to complement and enhance existing implants thereby improving the final outcome for patients. The basis for making an implant is a doctor’s request for making such an implant and a CT scan in order to precisely shape the bone that needs to be replaced. Such implants are created in close collaboration with the surgeon-operator with whom each individual feature is analyzed and coordinated. After the doctor agrees on the final design, the implant is produced by additive manufacturing technology, popularly called 3D print technology, which represents a revolution in the production of medical implants because of its speed, accuracy, and economy.


To Know more about Juniper Online Journal of Orthopedic & Orthoplastic Surgery

To Know more about our Juniper Publishers

Wednesday, May 8, 2024

Misidentification of Specimens Threatens the Integrity of Helminth Parasite Research - Juniper Publishers

Wildlife & Biodiversity - Juniper Publishers

Abstract

With the devaluing of the skills of morphological taxonomists misidentification of parasites appears to be increasing. In this paper examples from the acanthocephalan literature are reported. The need for morphological taxonomic expertise is emphasised including the importance of examining specimens.

Keywords: Helminth misidentification; Identification accuracy; Taxonomic expertise; Acanthocephalan; Amphibian; Fish; Parasite identification

Introduction

Of recent years, with the development of molecular techniques, the use of morphology in determining accurate identification of specimens has become undervalued. The knowledge and skills required for accurate morphological identification of individual worms is being disregarded and too much reliance is being placed on published host parasite lists and misused identification keys. As emphasised by Bush et al. [1], the result is a growing record of misidentified parasites in the literature, the consequence of which is to cast doubt on the reported findings. The absolute requirement for accuracy of specimen identification goes beyond the realm of formal taxonomy, affecting both the validity of phylogenetic and ecological analyses as well as the consideration of pathogenesis, zoonosis and control.

Although molecular data are becoming increasingly important in genus and species resolution, unless the samples are carefully characterised, their identity, as posted on GenBank, may be compromised. The correct identification of helminths requires satisfactory fixation and clearing for microscopic examination. Inadequately prepared specimens do not show characteristic morphology, making accurate identification difficult or impossible. Moreover, photographic images may not provide the detail necessary to make confident decisions. Another source of error, as reported by Bush et al. [1] is reliance on lists of published host records, which may be outdated or incomplete, for parasite identification.

Over the past few years, I have found several instances of Acanthocephala (thorny headed worms), incorrectly identified in the literature. These errors in acanthocephalan, and indeed of any other helminth parasite, identification have the potential to cause confusion at best or significant error in analysis at worst which may have serious consequences.

Corynosoma is a cosmopolitan acanthocephalan genus, largely parasitic in dolphins, seals and sea-lions. In the molecular analysis carried out as part of a study of the acanthocephalan Corynosoma hannae by Hernandez-Orts et al. [2] an isolate, registered on GenBank as Corynosoma australe by Garcia-Varela et al. [3] was found to have need mistakenly identified as C. hannae. The two species are clearly distinguished morphologically by the shape of the proboscis and the number and arrangement of hooks on the proboscis (the proboscis armature). Use of this sequence data, now known to be C. hannae, as that of C. australe in subsequent study of the genus will compromise any resulting analysis of geographic distribution, host-parasite relationships or infection parameters since all of these depend on accurate species identification.

Acanthocephalus ranae is an acanthocephalan parasite of amphibians found across Europe, including Turkey, with significant pathology described in host intestines [4]. Sakthivel and Gopalakrishnan [5] described seasonal variations, studied over 3 years, and pathological lesions caused by Acanthocephalus ranae in fish hosts from coastal locations in Tamil Nadu. These authors provided a brief description, drawings and photographs of the putative A. ranae as well as infection data, and histopathology and histochemistry descriptions. Unfortunately, although it can be clearly seen from the images that the specimens in question are not A. ranae, it is impossible to determine which species they might be. In this instance the unwary reader might conclude erroneously that the geographic range and host species of A. ranae have been extended from Europe and amphibians to India and fish. Furthermore, the inaccuracy of the parasite identification throws some doubt on the description of the pathology caused by these unknown acanthocephalans.

Conclusion

The discovery of misidentifications such as the two examples outlined above highlight the importance of developing taxonomic skills in order to undertake thorough morphological examination of specimens. Careful comparisons with previously identified material, if available from museum collections and published descriptions should be carried out before species identifications are made. The importance depositing specimens in publicly available collections for the purposes of comparison was emphasized by Hernandez-Orts et al. [6]. These authors also pointed out the importance of examining type specimens when possible as an aid to identification, because of the morphological data they exemplify. As emphasised by Bush et al. [1] erroneous information cascades are created if misidentifications are repeated in the literature, leading to mistaken conclusions. These authors proposed a set of guidelines for use by authors, editors and reviewers to minimise the likelihood of errors of identification. These guidelines include accessing the skills of morphological taxonomists, which I endorse.

To Know more about JOJ Wildlife & Biodiversity

To Know more about our Juniper Publishers


Monday, May 6, 2024

Skin Necrosis from a Spider Bite: A Case of Cutaneous Loxoscelia - Juniper Publishers

 Clinical & Medical Imaging - Juniper Publishers


   Let me send you a case of cutaneous loxoscelia. A 55-year-old woman from Tetouan, a region in northern Morocco, who has had diabetes for 6 years on insulin, reporting that she felt a sting in the right forearm while cleaning her home. The following day, she consulted for the sudden appearance of a pruritic and painless purplish erythematous plaque in the right forearm (Figure 1A) in a context of apyrexia and preservation of the general state , evolving a few hours later into an aspect of cockade centered by a bubble (Figure 1B), the patient was put under local care but the evolution was marked by the appearance of a necrotic plate surrounded by an inflammatory halo (Figure 2). A biological test made of an NFS, TP, TCA, liver enzymes were normal, and the bacteriological samples were sterile. The diagnosis of cutaneous loxoscelism was suspected given the context. The patient underwent surgical detersion of the necrosis with local care and directed scarring with suitable dressing.


The loxoscelisme is a serious form of poisoning by spider bites of the loxoscele kind very little described in the literature, the cutaneous attack is in the foreground with a skin necrosis in the extreme cases. Our case was one by supposed spider bite. These are small ubiquitous spiders that can be found around the Mediterranean where the species Loxosceles rufescens lives. The venom of loxosceles contains many enzymes, in particular sphingomyelinase D, which induces complement activation, hemolysis, platelet activation and vascular thrombosis. The positive diagnosis of osteo-osmosis ideally rests on the identification of the responsible spider, which is rarely done in practice. In our observation, the diagnosis of cutaneous loxoscelisme was considered very probable, in the absence of formal proof by capture of the spider. It is based on a set of arguments: biting or little painful bite, characteristic clinical presentation, elimination of differential diagnoses and presence of Loxosceles rufescens confirmed in the region (Mediterranean). The loxoscélisme is a pathology ignored by the practitioners thus under diagnosed, it must however be considered by the dermatologists like a possible cause of cutaneous necrosis.

To Know more about Current Trends in Clinical & Medical Imaging
Click here: https://juniperpublishers.com/ctcmi/index.php

To Know more about our Juniper Publishers
Click here: https://juniperpublishers.com/index.php


Thursday, May 2, 2024

Characterization of Aroma Active Compounds of Cumin (Cuminum cyminum L.) Seed Essential Oil - Juniper Publishers

 Bioequivalence & Bioavailability - Juniper Publishers

Abstract

Cumin (Cuminum cyminum L.) is one such most popular spice that is used as a culinary spice for their special aromatic effect. The flavor of cumin is judged by its volatile oil content. The advantage of use of volatile oil is that it is 100 times more concentrated then the spice powder and hence is required in a very less quantity. The essential oil is responsible for the characteristic cumin odor. In present study evaluation of fragrance and flavor profile in essential oil of cumin from the Algerian market (Algeria, Northwest Africa) has been identified. The essential oil from the seeds of Cuminum cyminum L. was isolated by hydro-distillation method and the chemical composition was determined by gas chromatography-mass spectrometry. Eighteen (18) components representing (91.10%) of the essential oil were identified. β-pinene (9.5%), γ-terpinene (10.0%), p-cymene (11.8%) and Cuminaldehyde (50.5%) were the major components. The essential oil was also subjected to measurement of the physicochemical properties; refractive index (20 °C): 1.48, density (20 °C): 0.91, alcohol solubility (80% v/v): 1.1, aldehyde percentage: 50%, acidity: 1.0, alcohol percentage: 3.5%, carbonyl index: 9.32 and steric index: 19.24. These results suggested that the Cuminum cyminum L. essential oil is a potential source of active ingredients for food, pharmaceutical and cosmetic industry.

Keywords: Spices; Cumin; Cuminum cyminum L.; Essential oil; GC-MS; Physicochemical properties

Abbreviations: GC-MS: Gas Chromatography-Mass Spectroscopy; MSD: Mass Selective Detector; ISO: International Organization for Standardization; French AFNOR: French Association of Normalization

Introduction

Since earliest times medicinal plants have played a vital role in the development and comfort of human civilization. Many of the plants have medicinal properties that reduce symptoms or prevent diseases [1]. Spices are widely used in the Mediterranean countries of North Africa and Southern Europe. They are also used for their flavors and aromas and for the sensations they produce. They can also be used as food colorants and antioxidants [2].

Originally from the Mediterranean area [3], Cuminum cyminum L. is an annual herbaceous plant which grows up to 15-50cm height somewhat angular and tends to droop under its own weight. It has a long, white root. The leaves are 5-10cm long, pinnate or bi pinnate, with thread-like leaflets and blue green in color and are finely divided, generally turned back at the ends. The leaves are highly dissected. Whitish-red flowers are on a compound umbel (arrangement of flowers looks like an umbrella). The fruit is an elongated, oval shaped schizocarp (an aggregate fruiting body which doesn’t break open naturally and has two single seeded units called mericarps). The fruits are similar to fennel seeds, when chewed has bitter and pungent taste. The fruit are thicker in the middle, compressed laterally about 5 inch-long, containing a single seed [4].

Although the seeds of cumin (Cuminum cyminum L.) are widely used as a spice for their distinctive aroma, they are also commonly used in traditional medicine to treat a variety of diseases. The literature presents ample evidence for the biomedical activities of cumin, which have generally been ascribed to its bioactive constituents such as terpenes, phenols, and flavonoids. Multiple studies made in the last decades validate its health beneficial effects particularly in diabetes, dyslipidemia, hypertension, respiratory disorders, inflammatory diseases, and cancer. Cumin seeds are nutritionally rich; they provide high amounts of fat (especially monounsaturated fat), protein, and dietary fiber. Vitamins B and E and several dietary minerals, especially iron, are also considerable in cumin seeds [5].

The Cumin oil is reported as a high antioxidant mainly due to the presence of monoterpene alcohols [6]. The presence of phytoestrogens in Cumin has been reported which related to its anti-osteoporotic effects. Methanol extract of Cumin showed a significant reduction in urinary calcium excretion and augmentation of calcium content and mechanical strength of bones in animals [7]. Furthermore, the aqueous extract of Cumin seeds indicated the protective effect against gentamycin-induced nephrotoxicity, which decreased the gentamycin-induced elevated levels of serum urea and enhanced the clearance of the drug [8].

Essential oils have become in recent years a matter of considerable economic importance, with a constantly growing market whose fields of application are directly related to human consumption. This is why essential oils are more and more controlled in order to verify the presence of certain toxic natural compounds, their natural origin or not, their source and the presence of certain compounds. active ingredients. The purpose of this study is to provide experimental data on the chemical composition and the physicochemical properties of cumin that could be considered suitable for application in foods and drugs.

Materials

Plant material and essential oil extraction

The seeds of the plant were used; the plant material was hydro- distilled for 90min using a Clevenger-type apparatus. (The extraction performed after a 4-hours maceration in 500ml of water). The essential oil obtained was then dehydrated over anhydrous sodium sulphate and stored in a refrigerator at 4 °C until use. The plant was identified by Dr. Hicham Boughendjioua at the Department of Natural Sciences, High School Professors Technological Education, Skikda (Algeria). The voucher specimen under the plant’s name deposited then in the herbarium.

GC-MS analysis

Gas chromatography-mass spectroscopy (GC-MS) analyses of essential oil samples were carried out on a Hewlett-Packard 6890N gas chromatograph coupled to a HP 5973 mass selective detector (MSD). A HP5 column (30m х 0.32mm film thickness 0.25μm) was used. The analysis was performed using the following temperature program: oven isotherm at 35 °C for 5 min then from 35 to 250 °C at 6 ºC/min. Helium was used as the carrier gas at 1ml/min flow rate. The injector and detector temperatures were held, respectively, at 250 ºC. Mass spectra were recorded with ionization energy of 70eV and interface temperature of 280 °C. The identification of the oil constituents was based on a comparison of their retention indices relative. Further identification was made by matching their recorded mass spectra with those stored in the NIST mass spectral library of the GC-MS data system.

Results and Discussion

Classification of cumin

The plant was classified according to APG system III, 2009 (Table 1) [9].

Modern Applications of Bioequivalence & Bioavailability

Essential oil yield

The extracted cumin essential oil has dark yellow color, with an odor hot, powerful and spicy. The percentage yield of essential oil was calculated as per Moawad et al. [10], it is calculated on the weight basis. The equation is as follows: Volatile oil (%) = (Weight of the volatile essential oil recovered in g x 100)/Weight of sample taken in g. Yield estimation studies indicate that the value of essential oil was: 3.66%.

Physicochemical properties

Essential oils must meet characteristics imposed by the laws of producing and exporting countries and by importing countries. These criteria are defined in international standards ISO (International Organization for Standardization) or French AFNOR (French Association of Normalization). Thus, the organoleptic and physical properties such as coloration, odor, refraction, solubility, flash point, but also chemical properties such as acid and ester indices are controlled [11]. Physicochemical properties of the essential oil obtained by hydro-distillation from Cumin seeds are summarized in Table 2.

Modern Applications of Bioequivalence & Bioavailability

Chemical composition

Due to the enormous amount of raw product used to make wholly natural essential oils, it is important to study the chemical composition of the volatile fraction once the essential oil is extracted. Essential oils are hydrophobic and concentrated liquids whose composition is complex. The best qualitative and quantitative identity card of an essential oil, however, remains its chromatographic profile, most of which is carried out in gas chromatography.

The chemical compositions of Cuminum cyminum L. essential oil are shown in Table 3, Figure 1. Eighteen (18) components representing 91.10% of the essential oil were identified. β-pinene (9.5%), γ-terpinene (10.0%), p-cymene (11.8%) and Cuminaldehyde (50.5%) were the major components.

Modern Applications of Bioequivalence & Bioavailability
Modern Applications of Bioequivalence & Bioavailability

The essential oil of the seeds of Cuminum cyminum L. from China was isolated by hydrodistillation in a yield of 3.8%. The chemical composition of the essential oil was examined by GC and GC-MS; 37 components, representing 97.97% of the oil, were identified. Cuminal (36.31%), cuminic alcohol (16.92%), γ-terpinene (11.14%), safranal (10.87%), p-cymene (9.85%) and β-pinene (7.75%) were the major components [12].

The main constituents at different harvesting time being cumin aldehyde (19.9-23.6%), p-mentha-1,3-dien-7-al (11.4-17.5%) and p-mentha-1,4-dien-7-al (13.9-16.9%). The results of GC and GC/MS analysis showed that the fruits should be harvested at the ripe stage for ideal volatile oil yield and composition [13].

GC and GC-MS analyses of the essential oil of Cuminum cyminum L. from the Alborz Mountain range of Iran revealed contained α-pinene (29.2%), limonene (21.7%), 1,8-cineole (18.1%), linalool (10.5%), and α-terpineole (3.17%) as the major compounds [14].

Cuminum cyminum L. seeds essential oil was isolated by hydrodistillation method and the chemical composition was determined by gas chromatography-mass spectrometry (GC/MS). The yield of the oil was found to be 3.0% (on dry weight basis). A total of twenty-six components, representing 96.7% of the oil were identified. Cuminaldehyde (49.4%), p-cymene (17.4%), β-pinene (6.3%), α-terpinen-7-al (6.8%), γ- terpinene (6.1%), p-cymen-7- ol (4.6%) and thymol (2.8%) were the major components in the oil [15].

Composition of the essential oil, which was obtained from the seeds of Cuminum cyminum L. collected from Ilam, was determined by GC-MS. In total, 25 components (83.36%) of essential oil were identified. Major constituents were Isobutyl isobutyrate (0.45%), α-thujene (0.5%), α-pinene (30.12%), sabinene (1.11%), myrcene (0.34%), γ-3-carene (0.21%), p-cymene (0.6%), limonene (10.11%), 1,8-cineole (11.54%), (E)-ocimene (0.1%), γ-terpinene (3.56%), terpinolene (0.32%), linalool (10.3%), α-campholenal (1.76%), terpinene-4-ol (0.6%), trans-carveole (0.7%), geraniol (1.0%), linalyl acetate (4.76%), α-terpinyl acetate (1.8%), neryl acetate (1%), methyl eugenol (0.2%), β-caryophyllene (0.42%), α-humulene (0.3%), spathulenol (0.56%) and humulene epoxide II (1%) [16].

The essential oil content in cumin samples from Serbian market ranged between 2.0 and 4.0%, with 22 identified compounds, among which the most abundant were cumin aldehyde, β-pinene, γ-terpinene, γ-terpinene-7 al and p-cymene. Post-distillation cumin seeds waste material that remained after the essential oil extraction contains total polyphenols of between 30.1 and 47.5 mg GAE/g dry extract, as estimated by the Folin Ciocalteu method. Hydroxybenzoic and hydroxycinnamic acids, as well as glycosides of flavonones and flavonoles, are the dominant polyphenols [17].

The major constituents of the essential oil from the cumin fruits under different conditions of storage were cumin aldehyde belonging to oxygenated monoterpenes and p-cymene, and β-pinene belonging to monoterpene hydrocarbons. Results indicated that at room temperature, the proportions of compounds with lower boiling temperatures such as β-pinene (1.57-10.03%) and p-cymene (14.93-24.9%) were decreased; however, cumin aldehyde (45.45-64.31%) increased during cumin oil storage [18].

The GC-MS analysis of cumin oil showed that eleven constituents were identified; seven hydrocarbon monoterpens (33.09%) and four oxygenated monoterpens (66.92%). The monoterpens were α-thujene (0.41%), α-pinene (0.90%), β-pinene (10.72%), β-myrcene (1.27%), α-phellandrene (1.18%), p-cymene (3.54%) and γ-terpinene (15.07%), and oxygenated monoterpens identified were cumin aldehyde (21.10%), carboxaldehyde (5.34%), 2-caren-10-al (17.74%) and cumin alcohol (22.65%) [19].

This deviation from the common chemo-types may be attributed to the effect of the factors that specifically affect the composition and yield of the essential oil, which include seasonal and maturity variation, geographical origin, genetic variation, growth stages, postharvest drying and storage [20-23].

Conclusion

Cumin (Cuminum cyminum L.) is the second most popular spice in the world, after black pepper, and used as a medicinal plant for aromatherapy and various illnesses. Determination of the physicochemical characteristics of the oil may establish by measurement of extraction yield, refractive index, density, carbonyl and steric indexes together with aldehyde, alcohol and acid contents.

In the chemical profiling, eighteen (18) components representing (91.10%) of the essential oil were identified, of which Cuminaldehyde with a concentration of (50.5%) was the main constituent, the physicochemical properties of the essential oil were also subjected to study (measurement).

Essential oils have become in recent years a matter of considerable economic importance, with a constantly growing market whose fields of application are directly related to human consumption. This is why essential oils are more and more controlled in order to verify the presence of certain natural toxic compounds, their natural or non-natural origin, their source and the presence of certain active compounds and even though the plant biomass a very promising source for the future, very little works has been done on the study of the organoleptic and physicochemical properties of aromatic fractions of cumin. Due to its chromatographic profile, the essential oil extracted by hydrodistillation of this plant has organoleptic and physicochemical properties very appreciated in perfumery and will be very coveted in the sector of the food, pharmaceutical and cosmetic industry.

To Know more about Modern Applications of Bioequivalence & Bioavailability
Click here: https://juniperpublishers.com/mabb/index.php

To Know more about our Juniper Publishers
Click here: https://juniperpublishers.com/index.php


Environment and Sustainability: New Possibility of Growth in Switzerland - Juniper Publishers

 Environmental Sciences & Natural Resources - Juniper Publishers Abstract Switzerland's tourism industry, while economically suc...