Clustered microcalcifications (MCCs) in mammograms are an important sign in the detection of breast cancer. Nevertheless, it is a complex and difficult task for radiologists to detect the clustered MCCs from the tissue background of mammograms only by naked eyes. This paper presents a prototype of a computer-aided detection system to automatically detect MCCs in mammograms. The detection algorithm mainly comprises three modules. The first module, called the mammogram pre-progressing module, inputs and digitizes mammograms into 8-bit images of size 2048x2048, normalizes the images, manually extracts the breast region from the background. The second module, called the feature extraction module, is achieved by using mixed features consisting of two wavelet features and two gray level statistical features. The wavelet features are generated by a five-level wavelet decomposition and reconstruction algorithm. The gray level statistical features used in this paper are median contrast and normalized gray level value. Finally, the third module, called the MCCs detection module, discovers MCCs in the images by using a classifier. This paper uses a three-layer artificial neural network (ANN) as a classifier to segment MCCs from the processing image. The ANN takes these four features generated in the second module as inputs. The output of the ANN corresponding to the true MCC pixels is then thresholded to segment out the true MCC pixels. One advantage of the designed system is that each module is a separate component that can be individually upgraded to improve the whole system. The algorithm is tested with a series of clinical mammograms. A sensitivity of more than 78% is obtained at a relatively low false-positive (FP) detection of 2.09 per image. The results are compared with the judgement of radiological experts, and they are very encouraging.
|