Deep Learning for Medical Image Analysis
CMPUT 605

General Information

Lecture Hours:	Weekly meeting Thursday from 14h00 to 16h00
Lecture Room:	Zoom Meeting
Instructor:	Prof. Pierre Boulanger
Office:	Athabasca Hall 411
Phone:	780-709-1260
Email:	pierreb@cs.ualberta.ca
Course Text:	Weekly readings

---

Course Description

The past twenty years of clinical applications of multimodal medical imaging (CT, MRI, US, PET/CT/MR, etc.) has revolutionized how medicine is practiced today by improving disease diagnostic and treatment. In the last decade, Deep Neural Networks (DNN) usage in this field has opened new doors to process those images allowing to perform automatic segmentation, multimodal sensor fusion and registration, and computer-aided diagnosis. This course will review the various DNN architectures found in the literature and then explore how they can be used in practical clinical applications. Course work includes homework, programming assignments, reading, and discussion of research papers, presentations, and a final project.

---

Course Topics

The course will closely follow the book by Zhou et al. entitled "Deep Learning for Image Analysis" from Academic Press. It will be complemented by relevant articles and extra notes provided in class. The course will cover:

· Basic Theory of Neural Networks and Deep Learning

· Various DNN Architectures

· Medical Image Segmentation Using DNN

· Medical Image Registration Using DNN

· Temporal Image Analysis Using DNN

· Computer-Aided Diagnostics and Disease Quantification Using DNN

· Software Tools for DNN

· Optimization of DNN

Course Prerequisites

This course is intended for graduate students with a medical imaging background, radiology, and related fields. It is primarily intended for students who need more insight into how DNN can be applied to medical image processing. The course assumes a background in programming and good knowledge of image processing. In addition to the medical relevance of the topics being covered, the exposition to the theoretical concepts will be made more concrete by a set of well-selected programming exercises. Familiarity and experience with Python and with vector-matrix calculus is a prerequisite for this course.

---

Course Evaluation

30%	Homework	There will be regular homework assignments
55%	Project	There will be one team-based semester project, culminating in a final report and a haptics "open-house" where the project will be demonstrated. Progress and checkpoints before the last due date will count toward the final grade.
15%	Presentation	The presentation is a detailed lecture on a topic related to haptics, done individually. The lecturer also prepares a short discussion or group activity for after the talk.

---

Lecture Notes

Lecture Date	Topics	Slides	Extras
Jan. 14	Introduction	Introduction.pptx NeuralNetworksinMedicalImagingApplications.pdf	Assignment 1 Assignment Due Jan. 29
Jan. 21	Basic Math of Deep Neural Networks	Understanding deep convolutional networks by Mallat Deep Neural Network Mathematical Mysteries for High Dimensional Learning
Jan. 28 to Feb. 11	Basic Math of Deep Neural Networks	Deep-Leaning-Basics.pptx Deep-Learning-Overview-MRI-2018.pdf Chapter 1 and 2 in Class Book	Overview of Deep Learning in Medical Imaging Assignment 2 Assignment Due Feb. 25
Feb. 18	U-Net and V-Net for Medical Image Segmentation	U-Net.pdf u-net-teaser.mp4 V-Net.pdf U2 Net	Project Proposal Due Feb. 25
Feb. 25	Deep Belief Networks: Restrictive Boltzman Machines and Autoencoder	Tutorial on Deep Belief Networks: RBM and Autoencoders TensorFlow Tutorial on Autoencoders	Assignment 3 Assignment Due Mar. 25
Mar. 4	A simple overview of RNN, LSTM and Attention Mechanism	RNN-LSTM-Tutorial
Mar. 11	Complex and Quaternion Neural Networks	Complex and Quaternion Neural Network Lecture Quaternion Neural Networks
Mar. 18	Fourier Neural Networks	Fourier CNN Implementations Fourier Neural Networks a Comparative Study A Fourier Domain Training Framework Tiled Fourier CNN Accelerating Convolutional Neural Network
Mar. 25	Multi-level Wavelet Neural Networks	MWCNN Paper Wavelet Neural Networks	Assignment 4 Assignment Due Apr. 15
Apr. 1	Medical Image Registration Using Neural Networks	CNN-based Medical Image Registration Deep learning in medical image registration: a survey
Apr. 8	Spiked Neural Networks	Spiked Neural Networks
Apr. 15	Solving ODE Using Neural Network	ODENet	Project Report Due Apr. 22

Grading Policy

• Your course grade is determined from the total points you receive from homework assignments, course projects, presentations, and class participation.
• Homework assignments are due at the beginning of class. You have two free late days to be used at any time in the course, except on presentations and projects. After that, no late assignments will be accepted.
• You are responsible for all information given in class verbally and/or in writing. Any information about the course on the web may be replaced by the information presented in class.
• Cooperative efforts at understanding the material and the assignments of the course are encouraged. However, you are required to submit only work that you have completed individually (except for the team project). Submitting any work that is not the result of a student's effort is considered cheating.

• ---