AI for Healthcare: How Can Hospitals Share Data to Build Better Models?

Abstract

The study has explored the importance of sharing data in hospitals for training Artificial Intelligence (AI) models. It has been observed that sharing data in the model training faces challenges such as ethical and privacy concerns, and the revelation of sensitive data of the patients. AI has been used in the detection of skin diseases and in the prediction of cancer from X-Ray images. However, limited data within a single hospital poses challenges and raises data privacy concerns. Investigation on this topic was found to be effective for improving the patient survival and recovery rates. The study has outlined several techniques to address the challenge of data privacy concerns while sharing data for AI training. This includes Federated Learning techniques, Homomorphic encryption, synthetic data generation and differential privacy.