| Abstract |
: |
Breast cancer is the most common cancer among the Indian women and it ranges from 25 to 31% of all cancers among Indian women. It is better to treat this dreadful disease at the earliest in order to save invaluable lives. For this sake, we developed a system that automatically detects and classifies the suspicious lesions present in the mammograms. The results are accurate because two levels of segmentations namely coarse and fine segmentation are employed. Coarse segmentation is done with the help of histogram based fuzzy c means technique, which is known for its accuracy, since it takes degree of truth and false into account. After obtaining the local sketch of the suspicious region, fine segmentation is applied in order to improve the rough representation of coarse segmentation and this is achieved by window based adaptive thresholding method. Finally, the outcome of fine segmentation is superimposed over the coarse segmentation to arrive at the perfect result. Then, the first order and run length features are extracted and the image is classified as normal, benign or malignant. Also, the type of lesion is identified by the maxvote algorithm and it proves 95% of accuracy. |
