An Intelligent Medical Diagnosis System

By: Babiker Mohamed – Digital Analyst

Think of a clinic in a faraway village, where lots of people stand in line to sign up, not knowing which doctor they will meet or what tests they will have. This might take a long time maybe all day and some can’t wait because of illness. In a world where technology is moving fast getting to care is still a big problem in poor areas. Smart computers can change this but current systems are often wrong. or don’t interact well especially with messy text like patient notes. This study wants to fix this issue by making a smart system that looks at symptoms from PDF files or text entries, asks helpful questions to clear up diagnoses and sends patients to the right doctors and tests with an eye on places with few resources. Here the benefits of al comes to solve these issues in the third world. country to help and assist the patients.

Although diagnostic systems like Ada Health or WebMD are around, they usually need serni structured input or fixed rule-based follow ups questions. There has been relatively limited work on extraction of symptoms from unstructured text (e.g., PDFs), as well as application of advanced approaches, like reinforcement learning, to learn dynamic follow up questions. In addition, there exist no systems for deployment in low-resource settings with poor access to medical resources. In this work we want to tap into these challenges, and for that we build on existing NLP and combine these with reinforcement learning (RI.) which we ground in practice.

This article was written by Osman Alhussein. Published with their permission.