ABSTRACT
This chapter introduces deep learning, in particular, convolutional neural networks for unsupervised and supervised image sematic segmentation. It presents the backpropagation algorithm for learning Convolutional neural networks (CNN) which extends the derivation of backpropagation in CNN based on an example with two convolutional layers. Medical imaging is a visual representation of the interior of the body for clinical analysis and medical intervention. Currently popular medical imaging techniques include abdominal ultrasound, contrast-enhanced computer tomography. The chapter develops statistical methods for imaging-genomic data analysis, in particular, using image segmentation as a framework. It examines causal inference as a general framework and powerful tool for imaging-genomic data analysis. Image segmentation plays an important role in prediction, diagnosis, treatment, and imaging-genomics data analysis. CNN are originated from application of neural networks to imaging data analysis and are designed to process the multiple array data.
