Computer-aided Diagnosis of Solid Breast Lesions Using an Ultrasonic Multi-feature Analysis Procedure

We have developed a family of quantitative descriptors in order to provide noninvasive, reliable means of distinguishing benign from malignant breast lesions. These include acoustic descriptors (“echogenicity,” “heterogeneity,” “shadowing”) and morphometric descriptors (“area,” “aspect ratio,” “border irregularity,” “margin definition”). These quantitative descriptors are designed to be independent of instrument properties and physician expertise. Our analysis included manual tracing of lesion boundaries and adjacent areas on grayscale images generated from RF data. To derive quantitative acoustic features, we computed spectral-parameter maps of radio-frequency (RF) echo signals (using a sliding-window Fourier analysis) of the lesion and adjacent areas. We quantified morphometric features by geometric and fractal analysis of traced lesion boundaries. Although no single parameter can reliably discriminate cancerous from non-cancerous breast lesions, multi-feature analysis provides excellent discrimination of cancerous and non-cancerous lesions. Our analysis of data acquired during routine ultrasonic examination of 130 biopsy-scheduled patients produced a receiver-operating characteristic (ROC) area under the curve (AUC) of 0.947±0.045. Lesion-margin definition, spiculation, and border irregularity were the most useful among the quantitative descriptors; some morphometric features (such as border irregularity) also were particularly effective in lesion classification. Our results are consistent with many of the Breast Imaging Reporting and Data System (BI-RADS) breast-lesion-classification criteria in use today.