JMIR Publications

ABSTRACT

Background:

MRI reports are challenging for patients to interpret and may potentially subject patients to unnecessary anxiety. The advent of advanced artificial intelligence (AI) language models such as GPT-4 hold promise in translating complex medical information into layman terms.

Objective:

To evaluate the accuracy, helpfulness, and readability of GPT-4 in explaining MRI reports of patients with thoracic or lumbar fracture.

Methods:

MRI reports of 20 patients presenting with thoracic or lumbar fracture were obtained. GPT-4o was prompted to explain the MRI report in layman’s terms. The generated explanations were then presented to 7 spine surgeons for evaluation. The MRI report text and GPT-4o explanations were then analyzed to grade the readability of the texts using Flesch Readability Score and Flesch-Kincaid Grade Level Scale.

Results:

The layman explanations provided by GPT-4o were found to be helpful by all surgeons in 17 cases, with 6 of 7 surgeons finding the information helpful in 3 cases. The GPT-4o explanations were considered accurate by all surgeons in 11 cases, with 6 surgeons considering the information accurate in 5 cases, and 4 or 5 surgeons considering the information accurate in 4 cases. Review of surgeon feedback on inaccuracies revealed that the radiology reports were often insufficiently detailed. The mean readability score of the MRI reports was significantly lower than the GPT-4o explanations (32.2 +/- 16.3 vs 53.9 +/- 8.1, p<0.001). The mean reading grade level score of the MRI reports trended higher compared to the GPT-4o explanations (11-12th grade vs 10-11th grade level, p=0.11).

Conclusions:

Overall helpfulness and readability ratings for AI-generated summaries of MRI reports were high, with few inaccuracies recorded. This study demonstrates the potential of GPT-4o as a valuable tool for enhancing patient comprehension of MRI report findings.