Statisical Methods in NeuroImage Analysis

Time: Every Friday 1:00-4:00PM between September 2 to December 9.
Place: Seoul National University Hosptial BLDG 001 ROOM 308. This is the first building you see from the gate near Hyewha subway station exit 3.

Course webpage: http://brainimaging.waisman.wisc.edu/~chung/neuro.processing



Multiscale reprsentation of the sulcal pattern of human brain projected onto a sphere.


Instructor

Moo K. Chung
Associate Professor of Statistics, Biostatistics and Medical Informatics
University of Wisconsin-Madison

Visiting Associate Professor of Brain and Cognitive Sciences
Seoul National University

Office Hour: Friday 4:00-5:00PM.
Email: mkchung@wisc.edu
Official Webpage: www.stat.wisc.edu/~mchung

 

Prerequisites

None. The course is self contained but learning curve is expected to be steep. Need an acess to MATLAB. You don't need to know how to use MATLAB. But you will be taught how to program in MATLAB


Target Audience

The course is designed for graduate and advanced undergraduate students, and researchers with strong quantitative skills.
No knowledge in any sort of image analysis and computation is required and all course materials will be self-contained. The course material is applicable to a wide variety of applications using brain images (MRI, DTI, PET, fMRI etc).


Course Aim

Present statistical techniques used in analyzing various brain images. A concise review of relevant methodological background will be presented. Basic concepts of key methods will be developed with considerable attention to analysis of real brain imaging data of various types and problems. Students and researchers should gain a deeper understanding of statistical methods used in brain imaging. Course projects will be designed to apply methods learned from classes to real data.


Course Evaluation

Students are required to submit a final research report and do an oral presentation at the end of the semester. Students can use their own medical images for the final project after consultation with the instructor. For students without their own data set, the final project topics and data set will be provided. A-grade sample final project reports can be found here.


The evaluation is based on the successful completion of a course project based on techniques covered in class. Students are required to submit an initial report containing preliminary proposal before the course drop deadline (10%), give 20-minute oral presentation (20%) and submit the final report (30%). There will be exams (40%).  In last two years, 16 students took similar courses and the grade distribution is as follows: 1 A, 8 A-, 5 B+, 2 B, 1 C.


Text book

Computational Neuroanatomy: The Methods. This is a preprint of book I am currently writing and it will be freely available to students in the
dropbox.


Course Outline

Various statistical issues in neuroimage processing and analysis will be addressed. The focus of the course is on the computational aspect of various statisical procedures used in neuroimages. MATLAB will be used as a language of instruction although students can do homework and project in any computer languages of their choice. The following topics will be covered:

Week 1: Introduction to statistical computing in MATLAB, General linear models and random field theory.
Week 2: Smoothness of fields, Type-I error. multiple comparision correction using random field theory.
Week 3: Machine learning. Support vector machines (SVM).
Week 4: Type-II error in correlated images. Sample size computation. Power analysis.

Week 5-6: Hilbert space methods and Karhunen-Loeve expansion in random fields.
Week 7-8: Linear models: general linaer models, mixed and fixed effect models.
Week 9-10: More multiple comparisions using FDR and permutation tests. Nonparametric test procedures.
Week 11: Time series analysis.
Week 12-13: Multivariate analysis.
Week 14: Student presentation.

Various applications will be covered in connection with these topics. Some topics overlap with the topics covered in the previous course I taught in 2009 and 2010. Old lecture notes given in 2010 can be here. Old lecture notes given in 2009 can be found here.


Additional Course Materials

Sample data and MATLAB codes given in 2009 and 2010 courses are here. Required old reading materials can be found here. Students are required to read at least 2 journal papers per lecture.