Triage of 2D Mammographic Images Using Multi-view Multi-task Convolutional Neural Networks

With an aging and growing population, the number of women receiving mammograms is increasing. However, existing techniques for autonomous diagnosis do not surpass a well-trained radiologist. Therefore, to reduce that require examination by radiologist, subject preserving diagnostic accuracy observed in current clinical practice, we develop Man Machine Mammography Oracle (MAMMO)—a decision support system capable determining whether its predicted diagnoses further radiologist examination. We first introduce novel multi-view convolutional neural network (CNN) trained using multi-task learning (MTL) diagnose predict radiological assessments known be associated with cancer. MTL improves performance triage efficiency while providing additional layer model interpretability. Furthermore, takes as input assessment predictions CNN determines or will most likely provide correct diagnosis. Results obtained on dataset over 7,000 patients show MAMMO reduced requiring reading 42.8% improving overall comparison readings done radiologists alone.