Residential Course Learning vs. AI Responses with ChatGPT: Comparative Study of Knowledge Acquisition

Abstract

<bold>Background:</bold> Artificial Intelligence (AI) is transforming medicine offering innovative solutions that improve patient care and operational efficiency. <bold>Aim:</bold> To compare the knowledge of thoracic ultrasound (TUS) acquired through a residential course composed of lectures and practical simulations with the knowledge provided by AI. <bold>Methods:</bold> Before and after the 4-day residential course, all participants underwent the administration of 10 MCQs. We then posed the same questions to ChatGPT. <bold>Results:</bold> Correct answers provided by participants before the course were 46% of the total, raising to 77% after the course (p=0.00195). We then asked ChatGPT to answer the same questions. ChatGPT answered correctly to 7 of the 10 questions (70%). Incorrect answers provided by the AI were related to ultrasound signs of pneumothorax, fluid bronchogram and pleural effusion. Before the course, 21% of partecipants answered the PNX question correctly, while 75% chose the same incorrect answer as ChatGPT. After the course, 77% answered correctly, and only 23% confirmed ChatGPT’s answer. For the fluid bronchogram question, 12% answered correctly before the course, while 44% chose ChatGPT’s incorrect answer. After the course, 59% answered correctly, and only 29% confirmed ChatGPT’s answer. Regarding pleural effusion, 31% answered correctly before the course, with 20% choosing ChatGPT’s incorrect answer. After the course, 61% answered correctly, and only 9% confirmed ChatGPT’s answer. <bold>Conclusions:</bold> ChatGPT could provide on-demand explanations, reinforce key concepts, and offer clarification on the interpretation of sonographic findings. However, hands-on experience and dedicated courses are essential in learning TUS.

Ähnliche Arbeiten