AI Based Multi-Modal Inferential System for Differential Diagnosis in Healthcare
For a system to be truly artificially intelligent, it should either think like a human; act like a human; think rationally or act rationally. To deliver quality care and to be accepted by us humans, it is highly important for an AI system to act humanly in healthcare. For a system to act humanly, Alan Turing proposes that it should have the following capabilities: (a). Natural Language Processing; (b). Knowledge representation; (c). Automated reasoning; (d). Machine learning; (e). Computer vision; and (f). Robotics. In the quest of developing an AI system that can act humanly, we have built a multi-modal inferential system for differential diagnosis that satisfies five of the six capabilities enlisted above as a part of the Turing’s test. At the core of our system is a novel Siamese Bayesian Network (SBN) for knowledge representation and automated reasoning for differential diagnosis. A Natural Language Processor (NLP), that extracts the context of a disease from the textual inputs by the patients, and a Computer vision system (CVS), that infers the findings from lab reports and medical images, feed into the SBN for better convergence of differential diagnosis. While the NLP uses a combination of SVM’s and LSTM’s for inference from text, the CVS employs mid level computer vision and deep mobile nets and inception nets for inference from images. In its current state, the system is capable of inferring from text, natural images (lab reports) and chest X-Rays. This talk summarizes the algirithms in use at delivering healthcare at scale in India.
An electronics engineer by education, Ajit has more than 20 years of expertise in creating products & building organizations across e-commerce, Consumer Internet, Mobile, Analytics, Integration and Platforms. He is an inventor on many patents in these domains and his current research includes Virtual Assistants, Explainable Artificial Intelligence and Knowledge Representation. In his previous avatar, Ajit was the CTO for Myntra, India's largest e-commerce store for fashion and lifestyle products. At Myntra, Ajit was responsible for leading a 400+ tech team and define strategy and direction for Consumer Applications, Supply Chain, Analytics and Data Science Platforms. Ajit started off in mobile technology at SAP, where he built products for offline and online mobile application development, domain programming languages for User Interfaces and Integration and API management.