Deep Learning for Medical Image Analysis

Chapter 14

Deep Learning Models for Classifying Mammogram Exams Containing Unregistered Multi-View Images and Segmentation Maps of Lesions¹

Gustavo Carneiro^⁎; Jacinto Nascimento^†; Andrew P. Bradley^{‡    ⁎}University of Adelaide, Adelaide, SA, Australia^†Instituto Superior Técnico, Lisbon, Portugal^‡University of Queensland, Brisbane, QLD, Australia

Abstract

In this chapter, we show two discoveries learned from the application of deep learning methods to the problem of classifying mammogram exams containing multi-view images and segmentation maps of breast lesions (i.e., masses and micro-calcifications). We first demonstrate the efficacy of pre-training a deep learning model using extremely large computer vision training sets, and then fine-tuning ...