Ultrasonic artificial intelligence-assisted diagnostic system for diagnosing medullary thyroid carcinoma / 中国医学影像技术

ABSTRACT

Objective To assess the effect of ultrasonic thyroid artificial intelligence(AI)-assisted diagnostic system(AI-assisted diagnostic system)for diagnosing medullary thyroid carcinoma(MTC)compared with different physicians and taken papillary thyroid carcinoma(PTC)as the controls.Methods Totally 63 MTC,70 PTC and 62 benign thyroid nodules confirmed by pathology were enrolled.AI-assisted diagnostic system was utilized to analyze thyroid nodules and identify the likelihood of malignancy,and the probability value threshold was set at ≥0.40.All thyroid nodules were retrospectively reviewed and categorized by 3 physicians(1 senior physician,1 attending physician and 1 junior physician)according to Chinese thyroid imaging reporting and data system(C-TIRADS).The efficacy of AI-assisted diagnostic system and physicians for diagnosing MTC and PTC were evaluated.Results AI-assisted diagnostic system showed lower sensitivity,specificity,positive predictive value,negative predictive value,accuracy,and area under the curve(AUC)for diagnosing MTC and PTC compared with physicians.Significant differences of AUC were found between senior physician and AI-assisted diagnostic system,as well as between attending physician and AI-assisted diagnostic system for diagnosing MTC and PTC(all P＜0.01),while no significant difference of AUC was between junior physicians and AI-assisted diagnostic system(both P＞0.05).The sensitivity,specificity,positive predictive value,negative predictive value,accuracy and AUC for AI-assisted diagnostic system for diagnosing MTC were all lower than those for diagnosing PTC,but the AUC was not significantly different(P＞0.05).Conclusion Ultrasonic thyroid AI-assisted diagnostic system had relatively high value for diagnosing MTC.