Assessment of Breast Lesions by the Kaiser Score for Differential Diagnosis on MRI: The Added Value of ADC and Machine Learning Modeling

How ADC could be combined with Kaiser score (KS) for the diagnosis of breast cancer was emerged as an interesting research area. We modified KS to KS+ based on the dichotomized ADC >1.4×10-3 mm2/s, and integrated KS and the continuous ADC values to build machine learning (ML) models for assessment. The diagnostic specificity of KS+ was higher than that of KS with a slightly degraded sensitivity. The AUCs of them were not significantly different. When the KS and the continuous ADC values were used to train ML models, the performance could be further improved while maintaining at a high sensitivity.