Evaluating the Qualitative and Quantitative Performance of Generative AI on Knowledge in Sports Medicine: The Case of GPT

This study evaluates the performance of generative AI, specifically the GPT 4 model, in sports medicine. It focuses on its ability to provide relevant and accurate information to healthcare professionals. Methodologically, the study included a series of tests in three categories: analysis of physiological curves, anatomical images and diagrams, and radiographs related to sports diseases, complemented by knowledge and clinical reasoning tests. GPT demonstrated strong skills, identifying physiological curve phases with an accuracy of 83.33%, demonstrating perfect accuracy in image and diagram analysis, and showing considerable accuracy in radiography interpretation despite minor errors. In knowledge tests, GPT achieved a 75.83% pass rate on advanced topics, underlining its applicability to specialist knowledge in sports medicine. Clinical reasoning tests further confirmed GPT’s ability to diagnose sports pathologies from symptom descriptions. Despite its potential as a supportive tool in sports medicine, the study identifies areas for improvement, such as data specificity and bias reduction. Improvements in GPT algorithms are recommended, along with greater collaboration between computer scientists and healthcare professionals to tailor these technologies to the unique needs of sports medicine. Clinical validation studies are suggested to ensure the efficacy and safety of these AI technologies in medical practice.