| Summary: | 碩士 === 國立成功大學 === 醫學工程研究所 === 87 === In this thesis, the statistical feature matrix and co-occurrence matrix were applied to quantify the texture of the ultrasonic image for the breast cancer, which often was described subjectively and qualitatively in clinical diagnosis. The breast ultrasonic images for thirty-three patients with palpable masses from May 1997 to May 1998 have been analyzed. The ultrasound imaging instrument is ATL HDI 3000. The parameters affecting image acquisition were kept consistent during image acquisition. The contrast and coarseness texture features were then computed from the statistical feature matrix, and the energy, entropy, contrast, variance, correlation, sum average and inverse different moment were derived from co-occurrence matrix. From the results, it is shown that most of the breast malignancies are more coarse, less contrast, more dark masses. On the contrary, the benign breast masses are less coarse, more contrast, and more bright. It is statistically significant that the risk of breast cancer is increased when the contrast and coarseness values of the nodule are below 12.46 and above 35.09, respectively, in statistical feature matrix. Along with the energy is above 0.00069, entropy below 7.5, contrast below 121.4, variance below 227.2, IDM above 0.122 and sum average below 42.7 in co-occurrence matrix.
|
|---|
