Analysis of ultrasonographic findings of breast masses by the Nagasawa Computer-Aided Diagnosis system

The Computer-Aided Diagnosis (CAD) software Nagasawa ver. 7.2 was developed by Professor Toru Nagasawa. The purpose of this study was to evaluate our decisions on ultrasonographic (US) findings using the CAD system. We analyzed 29 ultrasonograms (14 breast cancers and 15 fibroadenomas, diagnosed in Osaka City University Hospital). We categorized the US findings of the masses by the features of their shapes and borders (shape: regular, relatively irregular, irregular; border: smooth, relatively rough, rough). Then, we analyzed these ultrasonic categories with the CAD system. Each feature was compared with CAD parameters [shape: deformity index (DI), irregularity (Irr), marginal area index (MAI); border: entropy (Ent), fractal dimension (Fd), surface-length index (SL)]. The CAD parameters of DI, Irr (respectively, P = 0.0059), Ent (P = 0.0039), Fd (P = 0.0214), and SL (P = 0.0054) were significantly correlated with US categories. According to multiple comparisons, categories of "regular" and "relatively irregular," and "smooth" and "relatively rough," were in the same group. The CAD parameters indicated almost identical results with conventional US findings in breast masses. "Relatively irregular" and "relatively rough" can include "regular" and "smooth," respectively.