Practical Course:
Machine Learning in Medical Imaging

Nassir Navab, Diana Mateus, Tingying Peng, Shadi Albarqouni, Sailesh Conjeti, Christian Rupprecht, Nicola Rieke

Type: Master Practical Course Module IN2106
SWS: 6
ECTS: 10 Credits
Location: 03.13.10
Time: 10:00-12:00
Course Language: English

Announcements

• Final Presentation (7 mins.) will take place on Wednesday, 15th Feb. 2017 at 10 AM in the Seminar room 03.13.10
• Projects are assigned. Please contact your supervisor!
• Projects are uploaded (Missing ones will be up soon). Now, you need to send three preferences along with their project's ID to Shadi Albarqouni. Deadline to submit your preferences is Monday, 14th Nov. 2016
• The preliminary meeting will take place on 4th of July at 11:00 a.m. Seminar room 03.13.10
• Lectures and assignments are password protected
• To register for the course, please use the TUM-Matching-System

About the Course

The aim of the course is to provide the students with notions about various machine learning techniques. The course is subdivided in a lecture/excercises block and a project.

• The lectures will include topics in image processing and analysis, unsupervised methods and supervised methods. Each lecture will be followed by a practical hands-on exercise (e.g. implementation in Python).
• The topics of the projects will be distributed in the beginning of the semester. Each topic will be supervised by a different person. The projects are to be realised in couples. The programming language for the projects is to be decided in agreement with your supervisor.

Schedule

Date	Topic	By	Slides	Homework
19.10.2016	Introduction to Machine Learning and Python	Wadim Kehl	Intro Machine Learning Intro_Python
26.10.2016	Dimensionality reduction and clustering	Shadi Albarquoni	PCA PCA_Applications Clustering	Assignment data_code Have fun with faces dataset
02.11.2016	Vollversammlungen im Wintersemester 2016/17 (no lecture)
09.11.2016	Linear classification	Tingying Peng	Linear Classifier	Assignment SAheart (should work now)
16.11.2016	Support Vector Machines	Diana Mateus	Lecture SVMs	Assignment Two moons Digits
23.11.2016	Random Forests	Sailesh Conjeti	Lecture Random Forests Lecture Evaluation Measures	Assignment and Data
7.12.2016	Dies Academics (no lecture)
14.12.2016	Neural Networks and Autoencoders	Abhijit Guharoy	Lecture Neural Networks and Autoencoders	Assignment and Data
21.12.2016	Deep Learning	Christian Rupprecht	Lecture CNNs	Assignment (matlab + matconvnet) Assignment (python + tensorflow)
18.01.2017	Attestation	All tutors	There will be two sessions, please have a look at the schedule below
15.02.2016	Final Presentations

Evaluation

The grade will be split between weekly exercises associated to each lecture and a final project.

• The exercises will be evaluated during a mid-term attestation.
• The evaluation of the projects will include a final presentation

Projects

ID	Project	Tutor	Assigned Student	Material
1	Denoising of Magnetic Resonance Spectroscopy	Diana Mateus	Ruchik Yajnik	Slides
2	Cross-modality synthesis through style transfer	Diana Mateus	Ukrit Wattanavaekin	Slides
3	Hetero-Modal Medical Image Segmentation (HeMIS?)	Shadi Albarqouni	Emerson Tucker, Ahmed Elgazzar	Slides
4	Unsupervised Domain Adaptation	Shadi Albarqouni	Mario Anton	Slides
5	Spatial Transformer Networks	Sailesh Conjeti	Razieh Rezaei	Slides
6	Cutting out the middleman - Regressing cytological attributes sans segmentation	Sailesh Conjeti	Mihai Baltac	Slides
7	Super Resolution of Microscopy Images via CNNs (Histology)	Abhijit Guha Roy	Tamoor Ahmad	Slides
8	Deep Retinal Image Understanding	Abhijit Guha Roy	Adnan Naeem Husain	Slides
9	Adversarial Learning	Christian Rupprecht	Ignacio Sarasúa	Slides
10	Intra modality synthesis	Christian Rupprecht	Faisal Kalim	Slides
11	3D Cell Detection using CNN	Tingying Peng	Nina Stelzl	Slides
12	Online Learning with Random Forests for Tool Tracking	Nicola Rieke	Md Salman Rahman	Slides
13	Real-time Super-Resolution in Retinal Microsurgery (Ophthalmology)	Nicola Rieke	Christopher Kiwus	Slides
14	Retinal Vessel Segmentation	Nicola Rieke	Andrei Militaru	Slides

Attestation Schedule

Student	Time
10-11	Ahmad Tamoor Anton Mario Baltac Mihai Elgazzar Ahmed Husain Adnan Kalim Faisal
11-12	Kiwus Christopher Militaru Andrei Sarasua Ignacio Stelzl Nina Tucker Emerson Wattanavaekin Ukrit

Registration

• Interested students should attend the introductory meeting to enlist in the course.
• Students can only register through TUMOnline themselves if the maximum number of participants hasn't been reached.
• Maximum number of participants: 16.

Requirements

• *Strong knowledge of MATLAB, Python * or other programment language
• Background in image processing
• Exercises will be held in preferably in Python, althoug an introductory lecture on Python will be given previous knowledge would be advantageous.
• Basic knowledge of C++ is advantageous (in case you decide to use c++ for your final project): knowledge of main concepts of object oriented programming, basic software engineering, image processing libraries and linear algebra libraries.
• Basic knowledge of Python is advantageous (in case you decide to use python for your final project): numpy, scipy, scikit-learn.
• Basic knowledge of naive machine learning methods and terminology: knowledge of the meaning of the words classification, regression, data-set, cross-validation ...

FAQs

• Should we send the e-mail as a group? Or should each member of the group send it separately?
• Since we have rather enough projects, every student will be assigned to one project. However, there might be two projects with two members assigned by the course organizers. In short, send your preferences individually.
• Do we have to sort the options by preference, or are the three of them taken with the same priority?
• Yes, you need to sort them based on your preferences. We will do our best to match your first preference.
• For those projects that some knowledge on Deep Learning is required, is it enough with the one we will acquire during the lecture and the assignment or are we supposed to have some extra knowledge?
• Since the deep learning lecture will be held on 21st Dec., you won't have enough time to work on the project. Therefore, it is recommended to do some extra work.
• Is the tutor who is assigning each member of the group his/her particular tasks or are we splitting the work?
• In case if you are assigned with another student as one group, you need to discuss that with your tutor.