ABSTRACT
The growing significance and widespread adoption of artificial intelligence (AI) in healthcare have raised concerns from researchers, healthcare practitioners, and government bodies around the lack of transparency and responsibility towards data acquisition (source and quality) and ethical aspects from the inception of the algorithm to its field implementation. This inherent challenge was the backdrop against which the tank of the Government of India, frontier technologies supported a pilot study to validate the diagnostic performance of available AI algorithms developed in India for DR screening. Four Indian AI algorithms for DR detection agreed to participate. The research study considered ethics from inception (ethical clearance) to implementation (participant consent). The companies were requested to provide information regarding the training dataset encompassing patient demographics, race, ethnicity, camera specifications, pupil condition, quality of images, and the DR classification system. The efforts to get information proved futile as the AI companies shared minimal information. Therefore, an oversight body is required to develop and implement ethical risk-based frameworks for AI and technology to ensure AI’s large-scale, responsible adoption in healthcare settings.
