A Holistic Multi-Modal GenAI Healthcare System: Early Detection and Predictive Treatment Monitoring Using Cough Audio and Chest X-rays for Tuberculosis

Overview

Tuberculosis (TB), a bacterial infectious disease, is one of the top 10 causes of mortality worldwide in low- income countries, resulting in approximately 10 million new infections and 1.4 million deaths. Extending on my last year's research, I updated the ML model to include multi-modality (text and audio input) as well as additional feature extraction (mel -spectrograms) and data augmentation (IR -convolution). I created a mobile and web app to integrate the model, and the results are available within 15 sec onds. After preliminary testing, the model achieved an area under the receiving operator characteristic curve (AUROC) of 88%, surpassing the World Health Organization’s (WHO) requirements for screening tests. I also integrated my new research on chest radi ography (CXR) tools. I created a novel 2D convolutional neural network (2D-CNN) to identify and forecast subsequent incident TB using CXR. Predicting the risk of active TB long before symptoms enables preventive treatment that can be administered earlier. This research overviews the data exploration, development, training, and testing of various ML models, and an evaluation of the performance of the optimal ML model. After conducting some exploratory data analysis (EDA) on the CXRs and demographic information, I trained the model on the O2 high-performance cluster. I implemented Focal Loss (gamma=3.00 and alpha=0.95) and class weightage to counteract the high -class imbalance. The optimal 2D-CNN included a Gated Activation Unit (GAU) and a Multi -Head Self-Attention (heads=8), performing with a specificity of 93.7%, a sensitivity of 75.8%, and an AUC ROC of 83.7%, showcasing the strong potential for using CXRs in contact tracing. California Northern Plantsol: A Novel, Low-Cost Plant Disease Detection System Tejasveer Chugh Amador Valley High School, Pleasanton, CA American farmers are in crisis mode. More than 67% of the U.S. agricultural labor force has been lost in the last seventy-five years, making high-skilled agricultural workers difficult to find. Concurrently, increased production pressure is being put on Am erican agricultural systems, which play a key part in the world’s corn, soybean, and wheat production – staple crops for billions around the world. As a result of this coupling, farmers must take on a larger number of tasks, one of those being the detection of plant disease. This task is tedious, but also extremely important – a fact confirmed by interviews we personally conducted with farmers and agricultural experts in California. To manage plant disease, many farmers currently rely on manual approaches. Unfortunately, these are either inefficient or damaging to the environment. While automation is an option, existing solutions are either too expensive or limited to a specific environment or crop group. In response to this, we present Plantsol: a novel system that uses deep learning to identify 42 different diseases in 14 different crops with 93% accuracy, all for less than $50. The system is centered around an autonomous robot that can navigate crop rows and detect plant disease. A web application, which s yncs with the autonomous robot, and a mobile application, which provides on -device disease detection, treatment information, and access to global agricultural news, are also included. Plantsol is a breakthrough in the field of agriculture and may help solve the $220 billion problem of plant disease.