AlgoDoc

The Project

AlgoDoc is an Artificial Intelligence powered radiologist. It is a web-app that automatically diagnoses various Respiratory Diseases and Intracranial Hemorrhages from X-Rays and CT Scans. Healthcare facilities are scarce in rural areas across the globe , even more so in India. There continues to exist a huge shortage in the number of trained radiologists in India where there is only 1 radiologist for about 1,35,260 people. There are approximately 40 million diagnostic errors involving imaging annually worldwide, with the retrospective error rate among radiologic examinations being about 30%. AlgoDoc aims to empower people from rural and remote places to have a radiologist right in their pockets even if they can't afford a human radiologist by automatically diagnosing diseases in a very affordable way. AlgoDoc also serves as an assistant to radiologists, thereby minimising any error in diagnosis. AlgoDoc also provides a confirmation on the prediction from a verified human radiologist in-app. The prediction report will be sent to the user via mail as well for their future references. AlgoDoc provides users a Prediction Heat-Map, which makes its decisions interpretable.

View this project on YouTube

Team Comments

In the past year or two, everyone of us witnessed how the healthcare industry crumbled under the wrath of COVID-19. The healthcare industry, in countries around the globe were overwhelmed by the deadly pandemic. I therefore realised the pressing need to revolutionize the healthcare industry.

As the datasets are highly imbalanced, the use of normal Binary Cross-Entropy loss without any weights will result in a heavily biased model. This should be avoided under all costs and hence a Custom Weighted Binary Cross-Entropy loss function has been used to overcome the problem of Class Imbalance

1. Increase the accuracy of the AI to make its decisions much more reliable. 2. Partner with Hospitals to integrate AlgoDoc into their workflow 3. Getting approvals from various Quality Control councils (like FDA) by conducting various studies.