Comparing ChatGPT and DeepSeek for Generating Clinically Relevant Responses related to Physical Therapy

• 발행기관: KEMA학회
• 저자명: Jun-hee Kim
• 간행물 정보: 『Journal of Musculoskeletal Science and Technology』제9권 제1호, 9~18쪽, 전체 10쪽
• 주제분류: 의약학 > 재활의학
• 파일형태: PDF
• 발행일자: 2025.06.30

국문 초록

Background: Integrating large language models, such as ChatGPT, into healthcare has introduced new opportunities in medical education and clinical decision support. Recently, DeepSeek—an alternative artificial intelligence (AI) model optimized for computational efficiency—has emerged as a potential competitor to ChatGPT. However, the clinical accuracy and relevance of these models in physical therapy remain unclear. Purpose: This study aimed to compare ChatGPT and DeepSeek in generating responses relevant to musculoskeletal sciences and rehabilitation. Study design: A technical evaluation study Methods: A comparative analysis was conducted to evaluate ChatGPT and DeepSeek using six standardized questions related to musculoskeletal rehabilitation. Both models’ responses were evaluated by clinical expert using a 5-point scale based on six criteria including accuracy, coherence, fluency, reason-ing ability, justification, and medical suitability. Results: ChatGPT provided comprehensive and structured explanations with strong clinical rea-soning and justification, rendering it suitable for healthcare professionals. Meanwhile, DeepSeek generated concise, accessible responses optimized for quick understanding but lacked depth and justification. Although both models demonstrated good accuracy, ChatGPT’s responses were more suitable for professional use, whereas DeepSeek’s responses were more user-friendly for nonspecialists. Conclusions: ChatGPT exhibited superior clinical depth and justification, rendering it more appropriate for medical professionals and educators. DeepSeek’s computational efficiency and concise responses suggested its potential utility in patient education and telemedicine. Overall, a combined AI approach integrating depth and computational efficiency can enhance AI-driven healthcare applications. However, further validation in this regard is needed to optimize AI deployment in rehabilitation.

목차

INTRODUCTION

METHODS

RESULTS

DISCUSSION

CONCLUSIONS

REFERENCES

키워드

ChatGPT DeepSeek Large language models Physical therapy Rehabilitation