MARKING BREAST CANCER METHOD FOR MAMMOGRAMS USING A WATERSHED ALGORITHM †Kaori IIDA, †Atsushi HIGASHI, †Hiroki TSUKINOKISAWA,

Mammography image by breast radiograph have been widely used in a mass screening. On the other hand, interpretation imposes a heavy burden for doctor. Then, we propose that an automatic marking algorithm for detection of the suspected tumor of breast cancer in mammogram. This software supports doctor’s diagnosis by attracting attention. In developing the algorithm, we focused on gradient feature’s importance for tumor, so the proposed system is based on Watershed method which expands areas along the direction of gradient. According to the field test with doctor, the proposal algorithm has a capability to find cancer under 10 percents oversight