Methods In Medical Image Analysis - Spring 2013

BioE 2630 (Pitt) : 16-725 (CMU RI) : 18-791 (CMU ECE) : 42-735 (CMU BME)

Course Goals

To gain theoretical and practical skills in medical image analysis, including skills relevant to general image analysis.  The fundamentals of computational medical image analysis will be explored, leading to current research in applying geometry and statistics to segmentation, registration, visualization, and image understanding. Student will develop practical experience through projects using the new v4 of the National Library of Medicine Insight Toolkit ( ITK ), a popular open-source software library developed by a consortium of institutions including Carnegie Mellon University and the University of Pittsburgh.  In addition to image analysis, the course will include interaction with clinicians at UPMC via the Shadow Program.

Important Note about Videoing of Lectures

A few of the class lectures will be videoed for public distribution.  The camera will be in the back of the room, and we will attempt to point it over your heads. However, some students may at times be seen (usually only partially) by the camera. If you object to this, please see the instructor to discuss seating options to avoid ever being seen by the camera. Also, any audio in the classroom may be picked up and recorded by either the camera or the instructor's microphone. If on a particular occasion you want your voice removed from some part of the recording, then please inform the instructor after class that day.

Prerequisites

Knowledge of vector calculus,  basic probability, and C++ or python (most lectures will use C++).

NEW starting 2012:  ITKv4 includes a new simplified interface and many new features, several of which will be explored in the class.  Extensive expertise with C++ and templates is no longer necessary (but still helpful).

Requirements and Grading

• Method: The point system
  • Each question or problem in a quiz or homework is assigned a point value
  • Your cumulative grade for quizzes [or homeworks] is (the sum of points your earned on all quizzes) divided by (the sum of points you could have earned on all quizzes)
  • So, your course grade is equally affected whether you miss 1 point on a 3 point quiz, or you miss 1 point on a 10 point quiz. (This is not the case for the more typical "averaged percentages" method.)
• Attendance: Required
  • Checked using Quizzes
  • On some days the quiz may be signing your name on the roll.
  • Some days may not have any quiz at all (attendance not checked).
• Quizzes: 20%
  • Not present / not taken = 0
  • Lowest 2 are dropped (the 2 on which you missed the most points)
  • So, if you are gone for a week-long conference, then the two 0's won't count.
  • In case of extenuating circumstances requiring further absence, talk to me, but I must be fair to the class (i.e. harder on you).
• Homework: 30%
  • Your TA will help you before the assignment is due. When grading, he will not try to figure out a non-working mess of code!
  • Late policy: 0% for code that does not compile, run, and at least perform some part of the assignment. However, if you've made a reasonable effort in advance and have been working with the TA but still have not been able to get things to work, then we will be much more generous with partial credit and/or extra time, on a case-by-case basis.
  • Also, if you are using a different compiler than the TA, then you will be given a brief period of time to fix unforeseen cross-platform incompatibilities.
• Shadow Program: 10%
  • You submit 1 report for each clinical station you visit.
  • The first time you miss a station for which you are scheduled (without good reason), you may contact your instructor to reschedule for 50% credit for that station.
  • If you do not show up a second time, you will be removed from the Shadow Program, get a 0 for all subsequent stations, and your instructor will be very unhappy with you.
• Final Project: 40%
  • 15% presentation
  • 25% code
• Final Letter Grade
  • A final numeric score >= 90.0 guarantees at least an A-
  • A final numeric score >= 93.5 guarantees at least an A
  • Curving: Actual cut-offs may be lower, but the above are guaranteed.

Textbooks

• Required
  • Machine Vision, Wesley E. Snyder & Hairong Qi, ©2004, ISBN 978-0-521-16981-3 (paperback) or 978-0-521-83046-1 (hardback)
• Officially Recommended
  • Insight into Images: Principles and Practice for Segmentation, Registration and Image Analysis, Terry S. Yoo (Editor)
• Useful if you want to build your bookshelf
  • Handbook of Medical Imaging: Processing and Analysis, Isaac Bankman (Editor)
  • Computer Vision, Linda Shapiro
  • Algorithms for Image Processing and Computer Vision, J. R. Parker
  • Machine Vision: Theory, Algorithms, Practicalities, E. R. Davies

Schedule

Posted online at http://www.cs.cmu.edu/~galeotti/methods_course/
The lecture schedule (and some topics) are subject to change, depending in part on class interest and involvement.