A research study conducted by NIH and Weill Cornell Medicine, New York City trying to find the ease of implementing AI into the health care system. The study included the AI model and human physicians answering questions from the New England Journal of Medicine’s (NEJM) image challenge. In this challenge the participants need to answer real clinical images with short medical descriptions that include details about the presentation and symptoms that the patient came in with, they have to choose one answer among multiple choice answers.
The researchers asked nine physicians from various institutes from different medical specialties to answer the questionnaire of 207 questions in both “open book” and “closed book manner”. The AI model was prompted to answer the questions by choosing the correct diagnosis, describing the image, providing a summary of the needed medical explanation, and step-by-step reasoning.
Then the researchers found that the AI model was able to perform better in finding the correct diagnosis, even better than when the physicians answered with a closed book, but the physicians were able to answer better than the AI when allowed to use external sources (open book).
When the physicians were asked to evaluate the answers provided by the AI they concluded that even though the final diagnosis was chosen correctly, the AI model was not able to describe the reasoning for the diagnosis.
Even though the study could be a pioneer for the research on this topic it shed some light that when AI and human intelligence work hand in hand could help in improving and providing healthcare better to the world.
By Pothana Boyina Venkata Sai Vara Prasad
