ABSTRACT

This study aimed to verify whether the use of the Kaiser score can improve the diagnostic performance in breast magnetic resonance imaging (MRI) for suspicious lesions and avoid further invasive diagnostic approaches.

This retrospective study enrolled 97 patients who underwent breast MRI before undergoing breast biopsy or surgery. Evaluations were conducted on all MRI images individually by 2 radiologists using the Kaiser score. Neither radiologist had the knowledge of the final histopathological diagnosis. The ability of the Kaiser score in diagnosis was established via a receiver performing characteristic (ROC) analysis, which was measured by the area under the ROC curve (AUC). Youden index was used to define the optimal cutoff value. Kaiser score categories were dichotomized into positive (cutoff score > 4) and negative scores (cutoff score ≤ 4). Cohen’s kappa coefficient was used to analyze the inter-rater agreement.

Histopathology revealed 56 malignant and 41 benign lesions. The AUC for all lesions evaluated by the Kaiser score was 0.992 (95% CI: 0.981-1.0) and 0.958 (95% CI: 0.920-0.996) for 2 radiologists, respectively. Inter-rater agreement of the dichotomized Kaiser score was excellent (κ=0.894, P < .001). A total of 20 lesions (33.8%) previously categorized as BI-RADS 4 were reduced to BI-RADS 2/3 (19 benign lesions and 1 malignant lesion).

The Kaiser score is a valuable auxiliary diagnostic tool for improving the diagnostic ability of radiologists, whose experiences in breast MRI are diverse. In some cases, the application of the Kaiser score could possibly avoid unnecessary breast biopsies.