Progress 07/01/23 to 06/30/24
Outputs
Target Audience:The soybean cyst nematode (SCN) is the most damaging soybean pest in the United States and Canada. Sustainable soybean production is threatened by the narrow genetic base of resistance genes used in almost all SCN-resistant soybean varieties. SCN populations are building up reproduction on the commonly used resistance genes, and new varieties with unique resistance genes are desperately needed. Identifying resistant plants developed in soybean breeding programs requires largescale screening of plants in greenhouse experiments. The only way to determine if soybean lines are resistant to SCN is to manually dislodge adult cyst nematode females and cysts from roots with water and to count them. This process is tedious, unpleasant, time consuming, and labor intensive, thus expensive. To address thischallenge, we developed an automated instrument to imageadult SCN females and cysts from live soybean roots. Imaging modules and deep learning techniques wereincorporated for smarter identification of nematode cysts. Every month, 6 soybean plants as control and 6 soybean plants infected with SCNs were grown in the lab of Dr. Tylka and imaged in the Pandey lab. We collecetd an extensive database of SCN images. Various object detection algorithms were tested and traiined to recognize SCN cysts in live roots of plants grown in greenhouse. The final model has a resonably high accuracy (over 70% to 80% ) in detecting SCN cysts from live roots. Changes/Problems:
Nothing Reported
What opportunities for training and professional development has the project provided?Over 10 undergraduate students and one graduate student was supported by thisproject. Students worked on various aspects of image collection, model development, and data analysis. The project work helped thestudents to obtain research experience and secure jobs in industry. How have the results been disseminated to communities of interest?A object recognition model to identify and track microscale objects in videos has been disseminated to thepublic through an open-access publication in Scientific Reports, 2024. A Capstone Porject has demonstrated the project results to engineering students at the university. A soil processing company (AgSource) visited our laboratory to discuss patent licensing agreements based on our soil procesisng and imaging techniques. What do you plan to do during the next reporting period to accomplish the goals?
Nothing Reported
Impacts
What was accomplished under these goals?
In teh Tylka Lab, 12 soybean plants were grown every month for imaging puposes. Each plant root was imaged from four directions on a high--resolution scanner. The collected images were divided into training and test data. A state-of-art object recognition model was modified for our projectthat was originally develoepd by Meta for detecting people in big crowds. Using transfer learning, the model was trained to identify smaller objects (SCN cysts). The model works with reasonable accuracy in detecting SCN cysts.
Publications
- Type:
Journal Articles
Status:
Published
Year Published:
2024
Citation:
Kalwa, U., Park, Y., Kimber, M.J. et al. An automated, high-resolution phenotypic assay for adult Brugia malayi and microfilaria. Sci Rep 14, 13176 (2024). https://doi.org/10.1038/s41598-024-62692-x
|
Progress 07/01/20 to 06/30/24
Outputs
Target Audience:The soybean cyst nematode (SCN) is the most damaging soybean pest in the United States and Canada. Sustainable soybean production is threatened by the narrow genetic base of resistance genes used in almost all SCN-resistant soybean varieties. SCN populations are building up reproduction on the commonly used resistance genes, and new varieties with unique resistance genes are desperately needed. Identifying resistant plants developed in soybean breeding programs requires largescale screening of plants in greenhouse experiments. The only way to determine if soybean lines are resistant to SCN is to manually dislodge adult cyst nematode females and cysts from roots with water and to count them. This process is tedious, unpleasant, time consuming, and labor intensive, thus expensive. To alleviate the abovementioned critical bottleneck, we developed an automated instrument to load numerous root samples, prepare individual roots, and imageadult SCN females and cysts. Imaging modules were developed with deep learning for smarter identification of nematode females and cysts. In the Tyla Lab, we developed the strategies to grow soybean plants for 3 weeks in greenhouse every month. Strategies were develoepd to record images and videos of each live root using scanner and mobile phone from different directions. A mobile robotic platform was built to record images from a camera in a linear fashion and automatically collect the recorded images through USB connection. A number of deep learning models were trained to recognize SCN cysts from images first by manually labelling all the images and then testing the unlabelled images with the model developed by our team. The imaging and model were integrated to streamline the process of data collection and model inference. Changes/Problems:
Nothing Reported
What opportunities for training and professional development has the project provided?One graduate student and several undergraduate students in Electrical and Computer Engineering were involved inrecording images and videos, and developingalgorithms to be used for object detection for microscale objects. We had a Castone Projectof 5undergraduate students who labelled and tested the deep learning model. We have meetings and email communications with the team of the co-PI (Dr. Tylka) where we synchronize efforts to grow the plants every month and collect data on SCN cysts. Our team reviewed published literature on object detection algorithms between teh PIs to select the most efficient methods in image recording, labelling, and analysis. How have the results been disseminated to communities of interest?One journal publication has been published as an open access article with the full code of the object recognition model. Two undergraduate teams of senior students over the last two years developed posters to demonstrate the project. The project results have been shown and discussed with one soil testing company for possible licensing. What do you plan to do during the next reporting period to accomplish the goals?
Nothing Reported
Impacts
What was accomplished under these goals?
The major goal is to develop automated technologies to facilitate large-scale screening of root nematode (worm) infections in greenhouse experiments. We will build a robotic instrument to load numerous root samples, transport them on a train system, spray individual roots, and collect extracted adult soybean cyst nematode (SCN) females and cysts in catch sieves. Imaging modules developed with deep learning will be incorporated for smarter identification of nematode females and cysts. The three objectives are to: (i) design and build the hardware components, (ii) develop the root and adult female/cyst imaging modules, and (iii) integrate the entire system and conduct greenhouse experiments. In teh Tylka Lab, we have grown 12 soybean plants every month in soil having cysts and no cysts. The plant roots have been imaged every month to record images of the female cysts from four directions. An extensive database of the recorded images with SCN cysts have been created. Each image has been manullay labelled for the location of the female cysts using a deep learning program. The labelled images were used as training set to train a deep learning program to automatically count the total number of female cysts in an image.In addition, we have developed a robotic instrument to process soil samples, dislodge female cysts, and rupture the female cysts to release the eggs. The instrument was tested multiple times for workability and quality check. The instrument has been demonstrated to a soil testing company.
Publications
- Type:
Journal Articles
Status:
Published
Year Published:
2024
Citation:
Kalwa, U., Park, Y., Kimber, M.J. et al. An automated, high-resolution phenotypic assay for adult Brugia malayi and microfilaria. Sci Rep 14, 13176 (2024). https://doi.org/10.1038/s41598-024-62692-x
|
Progress 07/01/22 to 06/30/23
Outputs
Target Audience:Soybean cyst nematode (SCN) is the most damaging plant parasitic roundworm that infects the roots of soybean plants. Existing strategies to develop SCN resistant plant seeds have had reasonable success in reducing the infection rate by roundworms, but it is difficult to sustain the resistant traits against the SCN parasites across multiple generations. We are developing new robotic and imaging tools to conduct large-scale screening of SCN resistance in soybean plants grown in greenhouse experiments. Current methods to screen for SCN infection are laborious and tedious. We are building a mechanical system to load multiple plants on customized conveyor belts, bring them to a desired imaging module, and take a large number of images in three dimensions. The images will be processed by artificial intelligence to identify the female cysts on live rootsand decipher the traits with respect to their host plant. The automated instrument will produce data that can be integrated into current management strategies for SCN resistance. Changes/Problems:
Nothing Reported
What opportunities for training and professional development has the project provided?One graduate student and one undergraduate studentin Electrical Engineering have beenrecording images and videos, and investigating algorithms to be used for object detection. We have a senior designn team of four undergraduate students who evaluated thedifferent methods of labelling objects in images. We have meetings and email communications with the teamof the co-PI (Dr. Tylka) where we synchronize efforts to grow the plants every month and collect data on SCN cysts. Our teamreviewed published literature on object detection algorithmsand interacted with each other to select the most efficient methodsin image recording,labelling,and analysis. How have the results been disseminated to communities of interest?Our team was approached by a soil testing company in the Midwest to explore technology licencing options for our USPTO patent on soil processing and imaging to count the number of SCN eggs in soil. At present, negotiations are underway with this company for technology transfer. The Senior Design teampresented their results in their project review meeting to external reviewers in May 2023. What do you plan to do during the next reporting period to accomplish the goals?We will continue to grow soybean plants every month and build the image repository. The deep learning program to identify female cysts on live roots will be improved for better accuracy in detection. One source of error is the difficulty in recognizing the female cysts among the soil particles which have similar shape and color as the female cysts. We will try several image recognition algorithms to help differentiate the female cysts from the soil particles. Another task is to deploy the deep learning program on a portable smartphone which will be convenient for end users. We will explore methods to inncorporate light-weight computing approaches on smartphones to run the deep learning program.
Impacts
What was accomplished under these goals?
The major goal is to develop automated technologies to facilitate large-scale screening of root nematode (worm) infections in greenhouse experiments. We will build a robotic instrument to load numerous root samples, transport them on a train system, spray individual roots, and collect extracted adult soybean cyst nematode (SCN) females and cysts in catch sieves. Imaging modules developed with deep learning will be incorporated for smarter identification of nematode females and cysts. The three objectives are to: (i) design and build the hardware components, (ii) develop the root and adult female/cyst imaging modules, and (iii) integrate the entire system and conduct greenhouse experiments. We have grown 12 soybean plants every month in soil having cysts and no cysts. The plant roots have been imaged every month to record images of the female cysts from four directions. A database of the recorded images have been created. Each image has been labelled for the location of the female cysts using a deep learning program. The labelled images are being used as training set to train a deep learning program to automatically count the total number of female cysts in an image. In parallel, we trained a deep learning program using Amazon AWS to identify the female cysts in images. A comparison between the results of our deep learning program and the AWS program is being done to evaluate the performance of our methods. In addition, we have developed a robotic instrument to process soil samples, dislodgefemale cysts, and rupture the female cysts to relase the eggs. The instrument was tested multiple times for workability and quality check. The instrument has been demonstrated to a soil-testing company for potential technology transfer.
Publications
- Type:
Other
Status:
Published
Year Published:
2023
Citation:
Chris Cannon, Ethan Baranowski, Katherine Moretina, Matthew Kim, Santosh Pandey, "Soybean Cyst Parasite Detector", Senior Design Presentation, ECE Department, May 2023.
|
Progress 07/01/21 to 06/30/22
Outputs
Target Audience:Soybean cyst nematode (SCN) is the most damaging plant parasitic roundworm that infects the roots of soybean plants. Current strategies to develop SCN resistant plant seeds have had reasonable success in reducing the infection rate by roundworms, but it is difficult to sustain the resistant traits against the SCN parasites across multiple generations. We aredeveloping new robotic and imaging tools to conduct large-scale screening of SCN resistance in soybean plants grown in greenhouse experiments. Current methods to screen for SCN infection are laborious and tedious. We are building a mechanical system to load multiple plants on customized conveyor belts, bring them to a desired imaging module, and take a large number of images in three dimensions. The images will be processed by artificial intelligence to identify where the female cysts are and figure out their crytic traits with respect to their host plant. The automated instrument will produce data that can be integrated into current management strategies for SCN resistance. Changes/Problems:
Nothing Reported
What opportunities for training and professional development has the project provided?Three graduate students in Engineering are working on recording images and videos, and investigating algorithms to be used for objectdetection. We have meetings and email communications with the lab of the co-PI where we synchronize efforts to grow the plants every month and collect data on SCN cysts. The students review published literature on thsi topic and interact with each other to identify the challenges in data collection and analysis. How have the results been disseminated to communities of interest?Our published paper in Scientific Reports and USPTO patent is beingread by the community. We have received email inquiryfrom Syngenta and USDA National Lab in Tennessee to ask about the patent and apparatus developed. What do you plan to do during the next reporting period to accomplish the goals?We plan to sttudy the performance of different algorithms for object detection, and figure out which ones give the best results. We will continue to record images and videos of SCN roots to build a sizeable repository for model development. We plan to reach out to investigators working on SCN cysts to discuss our methods once we have a working algorithm for cyst detection.
Impacts
What was accomplished under these goals?
We have investigated a number of computer vision algorithms that recognize small objects from large images and video files. We have reserached other applications where small objects need to be identified from images or videos, including algorithms that work with images collected from drones. In most cases, there is a lack of consistent or coherent data files from the community. Furthermore, some available algorithms in the community are not sufficiently accurate or adaptable to our needs. As such, we are buildingare own data repositoryof root images and videos withh and without SCN cysts on them. The co-PI (Dr. Greg Tylka) provides us with a number of live plants grown in their greenhouse for 30 days each month, which are imaged in the Engineering Laboratory every month.We have imaged roots for over 7 months. WeuseAmazon Web Services for storing data and performing computer vision tasks. We are also building our own algorithms to recognize the SCN cysts from the live soybean roots.
Publications
- Type:
Other
Status:
Published
Year Published:
2021
Citation:
Santosh Pandey, Augustine Beeman, Leland E Harker, Jared P Jensen, Upender Kalwa, Taejoon Kong, Zach L Njus, Gregory L Tylka, Christopher M. Legner, Methods, apparatus, and systems to extract and quantify minute objects from soil or feces, including plant-parasitic nematode pests and their eggs in soil , US Patent 10,900,877, January 26, 2021.
- Type:
Theses/Dissertations
Status:
Published
Year Published:
2020
Citation:
Christopher Legner, Robotic instrument and automated imaging techniques to extract and count cysts and eggs of plant-parasitic nematodes from field soil, Iowa State University, PhD Dissertation
|
Progress 07/01/20 to 06/30/21
Outputs
Target Audience:Soybean cyst nematode (SCN) is the most damaging plant parasitic roundworm that infects the roots of soybean plants. Current strategies to develop SCN resistant plant seeds have had reasonable success in reducingthe infection rate by roundworms, but it is difficult to sustain the resistant traits against the SCN parasites across multiple generations. We are developing new tools to conduct large-scale screening of SCN resistance in soybean plants grown in greenhouse experiments. Current methods to screen for SCN infection are laborious and tedious. We are building a mechanical system to load multiple plants on customized conveyor belts, bring them to a desired imaging module, and take a large number of images in three dimensions. The images will be processedby artificial intelligence to identify where the female cysts are and figure out their crytic traits with respect to their host plant. The automated instrument will produce data that can be integrated into current management strategies for SCN resistance. Changes/Problems:
Nothing Reported
What opportunities for training and professional development has the project provided?Two undergraduate students worked remotely on surveying the various hardware options, discussing the methods for conductingmechanical functions, and identifying the available literature. One graduate student, Chris Legner, passed his Ph.D. defense in December 2020 and is working for a leading agri company in Ames, Iowa as a lab automation scientist. The PI has had several weekly meetinsg with students to train the studentson scientific writing for peer-reviewed publications. We have submitted 2 review papers based on our efforts, and they are under first revision. How have the results been disseminated to communities of interest?Our published paper on a soil processing instrument for soybean crops in Scientific Reports (2021) is accessible to anyone through the Open Access channel. The College of Engineering at ISU released a press release on our work published in this paper. What do you plan to do during the next reporting period to accomplish the goals?Next, we plan to start building different modules of the hardware setup to be able to capture images of plant roots. We have soybean plants growing in the lab of the co-PI, Dr. Greg Tylka, that will be used for thispurpose. We will establish a reliable method of capturing images of plant roots that has sufficient clarity and resolution to identify soybean cyst nematodes. A database will be set up to store the images. In parallel, we will investigate appropriate algorithms to process the root images.
Impacts
What was accomplished under these goals?
Our accomplishments were related to the design and configuration options of the hardware system, while learning about the options for large-scale image processing. We investigated methods of building thehardware platform that would comprise the conveyor belt, loading station, imaging modules, and root spraying system. Each component was searched online through Google and published papers to learn about the state-of-art techniques. From our research, we found there is no system in the public domain that is close to our proposed system, yet there are several handy designs that could be adapted. We identified a list of manufacturers that will be contacted to buy the mechanical parts. We have discussed the project with a lab technician whohandles the CNC machining of mechanical parts from CAD layout. We have built the CAD design files for the different hardware components, and are investigating various options to simplify the process with least number of parts. We have surveyed various imaging cameras and their features to understand which ones and which configurations would suit our purpose. While there are many simple options for imaging setups, our intent is to approach image capture from a perspective that has not been attempted before which provides a higher level of resolution in locating the female cysts on plant roots.In addition, we are looking into new ways to process images of plant roots with optimum digital resolution and yet having a large field of view.
Publications
- Type:
Journal Articles
Status:
Published
Year Published:
2021
Citation:
Legner, C.M., Tylka, G.L. & Pandey, S. Robotic agricultural instrument for automated extraction of nematode cysts and eggs from soil to improve integrated pest management. Scientific Reports 11, 3212 (2021).
https://doi.org/10.1038/s41598-021-82261-w
|
|