Abstract
Objective
This study aims to explore the application value of ChatGPT-4.0 in the ultrasonic image analysis of thyroid nodules, comparing its diagnostic efficacy and consistency with that of sonographers.
Methods
This is a prospective study based on real clinical scenarios. The study included 124 patients with thyroid nodules confirmed by pathology who underwent ultrasound examinations at Fujian Medical University Affiliated Second Hospital. A physician not involved in the study collected ultrasound images of the thyroid nodules, capturing three images for each nodule—the maximum cross-sectional, maximum longitudinal, and the section best representing the nodular characteristics—for ChatGPT-4.0 image analysis, classified according to the 2020 China Thyroid Nodule Malignancy Risk Stratification Guide (C-TIRADS). Two sonographers with different qualifications (a resident and an attending physician) independently performed the ultrasound examinations, also classifying the thyroid nodules according to the C-TIRADS guidelines. Using fine needle aspiration (FNA) biopsy or surgical pathology results as the gold standard, the consistency and diagnostic efficacy of ChatGPT-4.0 were compared with those of the sonographers.
Results
(1) ChatGPT-4.0 diagnosed thyroid nodules with a sensitivity of 86.2%, specificity of 60.0%, and AUC of 0.731, comparable to the resident's sensitivity of 85.1%, specificity of 66.7%, and AUC of 0.759 (p > 0.05), but lower than the attending physician's sensitivity of 97.9% and AUC of 0.889 (p < 0.05). (2) ChatGPT-4.0 showed good consistency with the resident in thyroid nodule classification (Kappa value = 0.729), but its consistency with pathological diagnosis was lower than that between the attending physician and the pathological diagnosis (Kappa values of 0.457 vs 0.816 respectively).
Conclusion
ChatGPT-4.0 has a certain clinical application value in the risk stratification diagnosis of thyroid nodules, comparable to the level of diagnosis by resident physicians.