Breast cancer diagnosis using image retrieval for different ultrasonic systems

This paper employs the image retrieval technique to classify breast tumors as benign or malignant lesions. We evaluated 600 ultrasound (US) images of pathologically proven solid breast nodules including 230 malignant and 370 benign tumors. The US images were acquired from four different ultrasound systems. Firstly, the physician located regions-of-interest (ROI) of ultrasound images. The textual features from ROI sub-image are utilized to classify breast tumors. The principal component analysis (PCA) is used to reduce the dimension of textual feature vector and then the image retrieval technique was utilized to differentiate between benign and malignant tumors. Historical cases can be directly added into the database and training of the diagnosis system again is not needed. The accuracy of the proposed computer-aided diagnosis (CAD) system was 91.2%, the sensitivity was 97.0% and the specificity was 87.6%. This system differentiates solid breast nodules with a relatively high accuracy in the different US systems and helps inexperienced operators avoid misdiagnosis.