Diagnostic Performance of Combining Apparent Diffusion Coefficient and Micro-calcifications to Kaiser Score in Evaluation of BI-RADS 4 Breast Lesions

A method with lower inter-reader variability for distinguishing benign lesions from BI-RADS 4 lesions needs to be developed to help avoiding unnecessary biopsies. The Kaiser score (KS) is a classification tree flowchart based on machine learning, which can guide clinical decision-making about breast imaging findings on breast MRI. It selects five morphological features and kinetics among 17 categorical diagnostic criteria to score each breast lesion and applies to both mass and non-mass enhancement (NME) lesions. The KS value ranges from 1 to 11, with increasing values indicating higher probability of malignancy. A biopsy is recommended, if the lesion score exceeds 4.

Ductal carcinoma in situ (DCIS), whose incidence has recently increased rapidly, only manifests as suspicious microcalcifications from time to time. Between 70% and 90% of DCIS diagnoses depend mainly on detecting microcalcifications on mammography.The morphology and distribution of microcalcifications are of great significance in distinguishing between benign and malignant micro calcifications. Given its confinement to the milk ducts, DCIS always presents as NME lesions and may be false-negative on MRI. A previous study suggested that the KS should be upgraded in the presence of suspicious mammographic microcalcifications to avoid missed DCIS diagnosis