Alternative Techniques for Breast Tumour Detection using Ultrasound

Ultrasound images are potentially invaluable for imaging internal organs and diseases. However, due to noise, they are still difficult to interpret. We apply and compare supervised machine learning approaches to train a model of lesions using features with unsupervised machine learning approaches to segment and detect tumours in breasts. Two synthetic and one real datasets are used in our experiments. The best system performance is achieved by Frost Filter with Quick Shift.