Artificial intelligence system and TI-RADS grading system in predicting thyroid cancer

Abstract

Objective To study the value of artificial intelligence (AI) systems and Thyroid Imaging Reporting and Data System (TI-RADS) in the diagnosis of benign and malignant thyroid tumors. Methods An AI system was developed, and a retrospective analysis was conducted on 500 patients who underwent thyroid surgery at Xuzhou Central Hospital from January 2020 to December 2021. Patient demographics, surgical information, AI grading indicators, TI-RADS classifications, and postoperative pathological findings were collected. The accuracy of AI and TI-RADS in predicting thyroid cancer was compared. Results Among the 500 patients, 390 were diagnosed with thyroid cancer and 110 were benign nodules. The AI achieved an accuracy of 88.5% in predicting thyroid cancer and 87.3% in predicting benign tumors, with an overall accuracy of 88.2%. The TI-RADS had a predictive accuracy of 80.3% for thyroid cancer and 80.9% for benign tumors, with an overall predictive accuracy of 80.4%. The AI group had higher accuracy rates for malignancy and overall prediction compared to the TI-RADS group (P＜0.05). The AI system had a higher accuracy rate for predicting papillary carcinoma compared to the TI-RADS (89.9% vs 81.2%, χ2=10.525，P＜0.05), while no statistically significant differences were found in predicting follicular carcinoma, medullary carcinoma, and undifferentiated carcinoma compared to TI-RADS (P＞0.05). Conclusion Both the AI system and TI-RADS have significant value in the diagnosis of thyroid cancer. The AI system demonstrates higher accuracy compared to the TI-RADS grading system, and similar results are observed in different subtypes of thyroid cancer.

Keywords

• thyroid cancer
• artificial intelligence
• thyroid imaging reporting and data system
• papillary carcinoma
• ultrasonic examination