Intelligent medical image analysis: a Deep Learning approach to breast cancer diagnosis
João Pedro Pereira Fontes · Portuguese National Funding Agency for Science, Research and Technology (RCAAP Project by FCT) · 2018
Once medical images were scanned and uploaded to a computer, researchers began to create automated medical imaging systems. From the 1970’s to the 1990’s, medical imaging was performed with sequential application of low-level pixel processing and mathematical modeling to solve specific tasks such as organ segmentation. At the end of the 1990’s, supervised techniques began to appear, where data extracted from the images were used to train models and classification systems. One example is the use of automated classifiers to build support systems for cancer detection and diagnosis. This pattern recognition and / or Machine Learning approach is still very popular and represented a shift from systems that were completely human-engineered to computer-trained systems with the use of specific (manually drawn) features and automatically extracted from the training data (example). The following step would be enabling the algorithms to directly learn characteristics of the pixels of the images. This is the basic concept of Deep Learning algorithms: multi-layered models that transform input data (images) into outputs (e.g. the presence or absence of pathological lesions or cancer). This study intends to present ways of using Deep Learning algorithms in the analysis of medical images, like the particular case of pathological lesions representative of breast cancer phenotypes.