Abstract
BACKGROUND: Suspicious breast lesions [Breast Imaging Reporting and Data System (BI-RADS) category 4 or 5] detected only by magnetic resonance imaging (MRI) and invisible on other initial imaging modalities (MRI-only lesions) are usually small and poorly characterized in previous literature, thus making diagnosis and management difficult. This study aimed to investigate the clinical significance of quantitative apparent diffusion coefficient (ADC) metrics derived from conventional diffusion-weighted imaging (DWI) on evaluating MRI-only lesions. METHODS: A total of 90 suspicious MRI-only lesions were evaluated, including 51 malignant and 39 benign lesions. Morphological and kinetic characteristics of all lesions (termed BI-RADS parameters) were described according to the BI-RADS lexicon on dynamic contrast-enhanced (DCE) imaging. Minimum, maximum, and mean ADC values (ADC(min), ADC(max), ADC(mean)) were obtained by measuring the ADC map of DWI. ADC(heterogeneity) was then obtained by the following formula: ADC(heterogeneity) = (ADC(max) - ADC(min))/ADC(mean). Diagnostic performance of these parameters was assessed and compared using the receiver operating characteristic (ROC) curve. RESULTS: Of the 90 MRI-only lesions, there were 45 masses and 45 non-mass lesions. Among BI-RADS parameters, only two different kinetic patterns were significantly different between benign and malignant groups (P=0.005 and P<0.001, respectively). The area under the ROC curve (AUC) of combined significant ADC parameters (ADC(min), ADC(mean), and ADC(max), all P≤0.001) was significantly higher than that of the two different kinetic patterns (P=0.006 for both). For MRI-only masses, only ADC(mean) and ADC(max), among all BI-RADS and ADC parameters, had diagnostic value (combined AUC =0.833). For non-mass lesions, size, distribution, ADC(min), and ADC(mean) were significantly different between benign and malignant groups (P=0.004, P<0.001, P=0.001, and P<0.001, respectively). In addition, ADC(mean) had the highest diagnostic performance among all ADC parameters, regardless of mass or non-mass (AUC =0.825 and 0.812, respectively). ADC(heterogeneity) showed no significant differences, no matter in mass or non-mass groups (P=0.62 and 0.43, respectively). CONCLUSIONS: In differentiating MRI-only suspicious lesions, quantitative ADC metrics generally performed better than BI-RADS parameters, and ADC(mean) is still the best ADC parameter to distinguish MRI-only lesions.