Performing Department
(N/A)
Non Technical Summary
Fresh produce grown hydroponically has recently been the subject of multiple recalls linked to foodborne illness outbreaks. The existing understanding of hydroponics (HP) and aquaponics (AP) operational procedures, practices, and conditions regarding food safety is insufficient to develop and implement effective food safety preventative control measures. This collaborative, integrated project will identify potential microbial contamination sources and the survival and persistence of human pathogens in the systems after contamination events and transfer to the final fresh produce. In addition, the project will validate the sampling size and frequency for contamination detection and develop AI-enabled biosensing to detect microbial contamination in HP and AP systems.Findings will help understand pathogen preference location for their persistence, factors influencing their survival and biofilm formation, and develop effective food safety risk management strategies. The project additionally seeks to enhance FAMU's capacity in AI-enabled biosensing, focusing on educating underrepresented students at FAMU and UMES and facilitating the creation of a cost-effective and rapid pathogen detection system. The project's outreach component includes educating small and medium-sized AP and HP growers in pre- and post-harvest food safety preventive control strategies.
Animal Health Component
15%
Research Effort Categories
Basic
80%
Applied
15%
Developmental
5%
Goals / Objectives
The main goal of this project is to evaluate microbial food safety risks and build capacity in AI-enabled Biosensing for Contamination Detection in Hydroponic (HP) and Aquaponic Systems (AP). Additionally, the project seeks to educate small and medium-sized AP and HP growers on pre-and post-harvest preventive control measures for ensuring food safety. The objectives of the project include: a) Determine a valid sampling method for contamination detection in HP and AP systems, b) Conduct Microbial Hazard Assessment in HP&AP systems, c) Develop AI-enabled sensing for contamination detection in HP&AP systems, d) Develop and deliver HP&AP production learning opportunities for students, faculties at FAMU & UMES e) Provide training, food safety plans, and standard operating procedures to current and potential HP&AP extension agents and growers.
Project Methods
Research: The project aims to determine a scientifically valid method for sampling to assess the hazard of pathogen contamination in recirculating nutrient solutions in HP and AP systems. The project will assess these factors using a combination of dosed inoculated challenge experiments, culture methods, and molecular techniques. The latter will provide insights into the microbial community diversity in nutrient solution samples compared to that on HP and AP plant roots from each location. Inoculated challenge studies will be conducted at three locations (FAMU, UMES, and USDA-ARS-BARC). Dosing will cover the range of 2 x 104, 9 x 104, and 2 x 105 CFU in the 2, 10, and 20 gal aliquots of HP and AP samples of recirculated nutrient solution. Water source and nutrient solutions (HP and AP), plant roots and shoots, and seedlings will be analyzed for microbial and physicochemical parameters three times a year for the first year. Risks resulting from splash events and human activities will also be evaluated, and sampling strategies will be developed to assess a variety of errant releases. The microbial community diversity of the samples will be determined using 16S rRNA sequencing and bioinformatics analysis.The project will develop an AI and optical imaging for contamination/rapid pathogen detection and evaluate its performance on real-world HP/AP water/ nutrient solutions with environmental background noise. It will produce a pre-trained dataset and develop software. FAMU & UMES HP&AP systems will be inoculated with rifampin-resistant, attenuated E. coli O157:H7 ATCC 25922 and Listeria innocua ATCC 33090. Furthermore, to ensure that the AI model can capture the diversity of E. coli and Listeria spp from the environmental samples, data for three strains of E. coli and three strains of Listeria innocua will be required. The water sample will be passed through a 50-micron filter to remove large debris, and the filtrate solution will be filtered from a 0.4-micron filter to concentrate bacteria from the filtrate. Bacterial microcolonies on TSA agar plates obtained from the filtered samples will be observed under phase contrast mode using an Olympus X71 inverted microscope with a 60×/0.7 Ph2 Air objective. Digital images (672 by 512 pixels with a pixel size of 107.5 nm) will be acquired by an ORCA-ER digital camera. Appropriate incubation time will be assessed to optimize microcolony growth and obtain sufficient morphological information on microcolonies. For each time point, 100 images will be obtained to calculate the average and standard deviation of the microcolony sizes. Microcolony size will be determined using MATLAB 2022a (The MathWorks, USA). To segment microcolony clusters from the agar background, the intensity of images will be normalized, and the images will be converted to a binary scale. The total pixel numbers of each microcolony cluster will be counted and reported as the microcolony size. YOLOv4 for microcolony identification and classification will carried out using MATLAB computer vision toolbox model for YOLOv4 object detection on a computer equipped with a 14-core central processing unit (CPU; Intel E5-2682 v4) and a graphics processing unit (GPU; NVIDIA Quadro P6000 24GB). The data set, containing ~ 400 images for each bacterial strain, will be collected through 10 replicate samples in 5 independent experiments. First, All the microcolonies in the data set will be labeled using the MATLAB image labeler. Each bacterial species will be annotated as a different class. Bounding boxes will be assigned to microcolonies, providing class and location information. Images will be shuffled randomly and split into a training data set (60%), a validation data set (10%), and a test data set (30%). The input image will be resized from 672 by 512 pixels to 608 by 608 pixels. Data augmentation will be performed on the training set with random horizontal flip, vertical flip, and rotation to improve training accuracy. CSP-darknet53-coco will be used as the network backbone for feature extraction. This detector has been trained on the COCO data set that consists of 80 different object categories. The pre-trained weights will be used as a starting point for transfer learning to increase the convergence rate. The validation set will be used to select the best parameters of the YOLOv4 detector during training. After the training, the performance of the YOLOv4 detector will be evaluated using the test set. The evaluation measures will include the intersection of over union (IoU), precision, recall, and mean average precision (mAP).Education: The undergraduate and graduate students receive theoretical and basic food safety skills through workshops and internships. Through hands-on instruction, they will learn firsthand how crucial it is to balance environmental, chemical, water quality, and microbiological aspects to provide the most favorable conditions for plants in HP systems and for plants and fish in AP systems. A workshop on AI-based contamination detection targeting undergraduate and graduate students will be conducted at FAMU in collaboration with UC Davis. The student interns will be mentored and taught HP&AP systems management and technologies, microbial and physical-parameter analysis, food safety practices, hazard control strategies, and potentially useful in fresh fruits and vegetable production systems. Intern students will participate in research projects on the safety of LGs grown in mid-size (research systems in greenhouse/high tunnels) and laboratory bench-scale systems. Student interns will learn HP and AP systems' engineering, operation, and food safety aspects. The students will also be trained in sample collection, processing, microbial microscopic imaging, and 16S rRNA sequencing for microbial community structure analysis.Extension/Outreach: The project aims to develop a curriculum and supplemental teaching material, for example, 1) videos, 2) hands-on demonstrations, 3) factsheets, and 4) standard operating procedures (SOPs). The project team will visit the commercial HP&AP farms before developing the course module, outlining learning objectives, and finalizing a timeline for training. The training will be provided mainly to extension agents and small to medium-sized HP & AP growers on HP&AP production/operation factors and systems inputs, human activities on microbial contamination, and potential preventive controls. The factsheets and instructional videos will be posted on the FAMU co-operative extension website every six months during the project.