The Prediction Model of High-Frequency Ultrasound Combined with Artificial Intelligence-Assisted Scoring System Improved the Diagnosis of Sclerosing Adenosis and Early Breast Cancer.

Abstract

Objective: The study aimed to apply an artificial intelligence (AI)-assisted scoring system, and improve the diagnostic efficiency of Sclerosing adenosis and early breast cancer.

Methods: This study retrospectively collected adenopathy patients (156 cases) and early breast cancer patients (150 cases) in Henan Provincial People's Hospital from August 2020 to April 2023.

Results: The area under the curve of the model constructed by clinical ultrasound features and combined AI features to predict and identify the two in the training group was 0.89 and 0.94, respectively. The combined AI model with the best performance (training AUC, 0.94, 95% CI, 0.91-0.97 and validation AUC, 0.95, 95% CI, 0.90-0.99) was superior to the clinical ultrasound feature model, and the decision curve also showed that the clinical ultrasound combined with AI Nomogram had good clinical practicability. In the training group, the AUC of the sonographer and AI in differential diagnosis was 0.67(95% CI, 0.62-0.71) and 0.89(95% CI, 0.84-0.93), respectively, and the sonographer's assessment showed better sensitivity (1.00 VS 0.73), but AI showed a higher accuracy rate (0.66 VS 0.80).

Conclusion: Age, lesion size, burr, blood flow, and AI risk score are independent predictors of sclerosing adenosis and early breast cancer. The combined clinical ultrasound feature and AI model are correlated with AI risk score, US routine features, and clinical data, superior to the clinical ultrasound model and BI-RADS grading, and have good diagnostic performance, which can provide clinicians with a more effective diagnostic tool.