Abstract
BACKGROUND: Patient comprehension of spine MRI reports remains a significant challenge, potentially affecting healthcare engagement and outcomes. Artificial Intelligence (AI) may offer a solution for interpreting complex medical terminology into layman's terms language. OBJECTIVE: To evaluate the effectiveness of AI-based interpretation of spine MRI reports in improving patient comprehension and satisfaction. METHODS: A prospective, single-center survey study was conducted at a single institution's multidisciplinary pain and spine clinics from May 2024 to November 2024, enrolling 102 adult patients scheduled for spine MRI. Imaging reports were interpreted using a single AI-based Large Language Model (LLM) that is securely operated within the hospital's network, with interpretations independently reviewed by healthcare providers and research coordinators. A board-certified neuroradiologist evaluated the accuracy of AI interpretations using a standardized 5-point scale. We analyzed survey responses from participants who received both their original MRI reports and AI-interpreted versions, comparing comprehension, clarity, engagement, and satisfaction. RESULTS: Participants reported higher comprehension with AI-interpreted MRI reports versus original radiology reports (8.50 ± 1.91 vs 6.56 ± 2.42; P < .001). AI interpretations received superior scores for clarity (8.57 ± 1.79 vs 6.96 ± 2.12; P < .001), understanding of medical conditions (7.75 ± 2.18 vs 6.27 ± 2.28; P < .001), and healthcare engagement (8.35 ± 2.00 vs 6.78 ± 2.48; P < .001). Accuracy assessment showed that 82.4 % of AI interpretations achieved high-quality ratings (≥4) [95 % CI: 69.7%-90.4 %], while 92.2 % were rated acceptable (≥3). Most participants (54.0 %) assigned the highest possible recommendation scores to AI interpretation. No significant differences were found between age groups and gender. CONCLUSIONS: AI-based interpretation of spine MRI reports significantly improved patient comprehension and satisfaction. Despite the promise of rapidly evolving AI-based technologies, a considerable percentage of AI interpretations were deemed to be inaccurate, warranting the need for further research.