An approach based on integrated solution for semi-automatic breast diseases investigation

Currently, breast echography is the most useful method capable of cheaper and accurate investigation of the breast diseases. The radiologists usually diagnose the benign versus malignant situations based on a competent visual inspection of the breast ultrasound (US) images. In the current study, we propose an original and efficient algorithm and a stand-alone Computer Aided Diagnosis application able to automatically diagnose and classify the breast lesions. The algorithm is based on the image pre-processing methods, image features, shape and orientation characterization and artificial neural network technique. Firstly, the input breast ultrasound image is classified as healthy or diseased based on a study of the mean gray intensity differentiation between the biological objects into image. Then, using the artificial neural network techniques the algorithm differentiates between malignant and benign cases. The validation step used three sets of ultrasound images: 20 healthy breast ultrasound images, 20 breast cyst images and 20 breast cancer images.