DEEP LEARNING FOR SYMPTOM-TO-DISEASE TRIAGE TO IMPROVE DIAGNOSIS IN RURAL AND UNDERSERVED COMMUNITIES
TL;DR
This work aimed to close this essential diagnostic gap by developing a practical, machine learning-driven system for symptom-to-disease prediction, and validated its accuracy for delivering basic, initial medical guidance and supporting early triage decisions for patients.
Abstract
Many people in rural and underdeveloped places continue to face significant chal-lenges in accessing dependable, qualified medical advice. When professional help is unavailable, patients may be forced to rely on traditional home remedies or local health myths, delaying the prompt and precise diagnosis required for effec-tive treatment. We aimed to close this essential diagnostic gap by developing a practical, machine learning-driven system for symptom-to-disease prediction. Our approach is simple; it takes a user’s reported symptoms and immediately generates a prioritized list of the top five most probable matching illnesses. We rigorously tested the solution using a variety of machine learning and deep learning techniques, including Random Forest, Decision Tree, Support Vector Machines (SVM), and a Deep Neural Network. Our final, optimized model achieved a robust 90% accuracy on a public dataset of disease and symptom specifications. Crucially, a clinical expert in homeopathy reviewed our model’s output and validated its accuracy for delivering basic, initial medical guidance and supporting early triage decisions for patients.