Electrogastrography and Personalized Transcutaneous Electrical Nerve Stimulation for Noninvasive, Lost-Cost Diagnosis and Treatment of Gastroparesis

Overview

Gastroparesis, characterized by slowed or absent gastric contractions, affects 1.8% of the population, presents with severe nausea, abdominal pain, a high mortality rate, and incurs high treatment costs. The condition poses diagnostic challenges due to its symptomatic ambiguity, and current treatments are ineffective, invasive, and costly. This study introduces an approach that combines transcutaneous monitoring of gastric contractions using an electrogastrogram (EGG) with personalized transcutaneous electrical nerve stimulation (TENS) to diagnose and treat gastroparesis. A dual-model approach customizes TENS parameters based on individual body mass indices (BMIs) for optimal gastric contraction stimulation, specifically targeting the PC6 forearm acupoint to enhance gastric motility. Using computational modeling, a COMSOL Multiphysics finite element model simulates the distribution of current in tissue following stimulation, and a NEURON Simulation Environment model assesses nerve activation based on this extracellular current distribution. Optimal parameters are the minimal intensity settings that activate 100% of nerve axons. The EGG setup was validated through phantom testing, and the simulations by comparing their results against published data. The method was evaluated in an in vivo study. The participants all had BMIs in the 20 - 25th percentile, and received optimized pulse parameters: 18 milliAmperes amplitude, 400 microsecond duration, and 30 Hz frequency. After treatment, gastric contraction frequency immediately rose by an average of 35.39% (p<0.01). Remarkably, a 30.35% increase in baseline contraction frequency was maintained 48 hours post -treatment (p<0.01), as evidenced by EGG, highlighting the therapy's prolonged efficacy. At $1053.49 this method offers an affordable, effective alternative for gastroparesis management. AR Surgical Navigation with Real-Time Carotid Artery Distance Monitoring and Tissue Visualization in Endoscopic Endonasal Surgeries using the Microsoft Hololens Kavin Ramadoss Sunset High School, Portland, OR The intricate nature of endoscopic endonasal skull base surgeries, where surgeons reach the brain through the nasal cavity and sinuses, necessitates precise navigation to avoid inadvertent encounters with the delicate carotid arteries, which can precipitate severe neurological consequences. This is why an innovative system that combines imaging modalities with a distance calculation algorithm is being built to solve this critical problem. The program enables real-time visualization of the surgical field, allowing surgeons to have sight of the patient's critical tissues and organs through a camera embedded in their instruments, coupled with a continuous assessment of the range and separation between the surgeon's instruments and the carotid arteries. The inte rface is designed to issue immediate warnings when the calculated distance approaches or breaches a predetermined safety threshold, thus equipping surgeons with invaluable, instantaneous feedback. The surgeons wear the Microsoft Hololens, an augmented real ity headset that allows them to see holograms displaying the distance and their surroundings. Furthermore, a custom convolutional neural network (CNN) was developed to extract intraluminal carotid artery models reliably from standard preoperative scans. By mimicking human visual processing, the algorithm achieves segmentation accuracy at around 91 \%. Patient-specific models then become projected holographically onto the operative scene through the Microsoft Hololens. This comprehensive program enables tremendous change in the medical field, causing surgeries to be accomplished quicker and significantly improving the mortality rate. Dual Machine Learning Architecture-Based Robotic Solution for Phytophthora Infestans Management and Mapping Ashank Shah Sunset High School, Portland, OR Phytophthora Infestans, generally referred to as potato blight, is a widespread, highly-infectious pathogenic disease among potato plants. Without immediate removal upon identification, its consequences on produce health are devastating. An average annual damage of $6.7 billion necessitates manual, systematic inspections for farmers, however, this process is labor-intensive and unsustainable. With current automated solutions being negligible, the implementation of an innovative, autonomous solution for potato blight management is crucial. Hence, a multi -faceted approach utilizing machine -learning algori thms, robotic automation, and mapping is developed. A robot utilizes a compact caterpillar drivetrain to navigate through rough, narrow terrain. To accurately identify blight, a webcam captures and parses image input through two machine-learning algorithms. The first is an object detection algorithm that locates individual potato leaves. Each respective leaf is subsequently passed into a Convolutional Neural Network that detects blight. Model optimization techniques including data augmentation resulted in a CNN accuracy of 99.88% and an RCNN mAP of 67.8%. To store blighted locations, a human -readable map of the crop rows is generated, highlighting blight-affected plants. Farmers are provided with updates of their crops' health via autonomous emails. A blight ed crop row was simulated through strategically placed images containing healthy and blight-infected plants in realistic terrain conditions in order to evaluate performance. The robot was successfully able to identify every instance of blight through multi ple trials. The promising results indicate that large-scale implementation and repurposing of this technology to combat a variety of plant diseases are feasible approaches in working towards universal crop-security. Pennsylvania Creating Digital Eyes for Visually Impaired Using Sensor Fusion and Stereo Vision Brandon Cai Parkland High School, Allentown, PA This paper demonstrates "Digital Eyes” as a wearable navigation aide mounted under the visor of a cap that can both direct a visually impaired user towards a destination using GPS and scan the walkpath for tripping hazards along the way. Inspired by autono mous driving technologies, a stereo vision camera is used to generate the 3D terrain model of the road, and a computer vision algorithm abstracts obstacle info from >300,000 points per camera frame. Through sensor fusion with 3-axis accelerometer, the system can automatically compensate for the fluctuation in camera angle and height as a user moves and turns the head by transforming the domain model to the ground frame. Despite this complexity, the “Digital Eyes” achieved a 4 frames -per-second refresh rate with all calculations completed within 0.25s including the Random Sampling Consensus (RANSAC) based ground plane detection, cluster scan -based object isolation, Rotating Caliper based Oriented Bonding Box (OBB) generation, and tripping hazard in path voice alert scripting with quantitative locations and dimensions info. Text-to-speech based alert enables users to “see with ears” and announce current location using reverse geocoding from GPS coordinates to addresses. The innovation of this project in making high resolution and fast scanning possible with limited computation power was adopting Robotic Vision’s techniques to focus on shape and sizes instead of following the beaten path of CNN based image recognition requiring significantly higher computation power and cost. This focus on the fundamental of tripping hazard detection makes this wearable assistive device affordable and practical.

Competition history

• JSHS 2024 Category not listed

Resources

• National JSHS abstract book ↗

Related projects

Closest projects by meaning, across every fair and year in the corpus.