Detecting and classifying lesions in mammograms with Deep Learning

摘要: In the last two decades Computer Aided Diagnostics (CAD) systems were developed to help radiologists analyze screening mammograms. The benefits of current CAD technologies appear be contradictory and they should improved ultimately considered useful. Since 2012 deep convolutional neural networks (CNN) have been a tremendous success in image recognition, reaching human performance. These methods greatly surpassed traditional approaches, which are similar currently used solutions. Deep CNN-s potential revolutionize medical analysis. We propose system based on one most successful object detection frameworks, Faster R-CNN. detects classifies malignant or benign lesions mammogram without any intervention. proposed method sets state art classification performance public INbreast database, AUC = 0.95 . approach described here has achieved 2nd place Digital Mammography DREAM Challenge with 0.85 When as detector, reaches high sensitivity very few false positive marks per dataset. Source code, trained model an OsiriX plugin availaible online at this https URL