October 27, 2006
Mini-conference on Medicine, Computer Science and Mathematics
Talks will begin at 3:30 PM in A226
Mark Wachowiak (Nipissing University)
Title: "Numerical Global Optimization in
Medical Image Registration"
Computation is now widely recognized as one of the three pillars of modern science, along with theory and experiment. As a major component of computational science, numerical optimization plays an increasingly significant role in all scientific endeavors. However, many important objective functions are non-convex, irregular, and their derivatives are not available or cannot be easily computed. Furthermore, the high computational cost of many global methods precludes their use in time critical-applications.
In this talk, a new, parallel approach to global optimization is presented for registering (geometrically aligning) medical images. Registration plays a vital role in image guidance of minimally invasive surgery and therapy, especially in cardiac procedures, where endoscopic images and real-time MRI and ultrasound scans must be quickly matched to pre-procedural 3D volumes. The intrinsic parallelism of two recent deterministic, derivative-free methods is exploited. Specifically, a global Lipschitzian approach is employed for balancing global and local search, and a simplex-based technique is used for local refinement. Experimental results on cardiac images with large morphology differences indicate a significantly higher success rate compared to local gold-standard methods, with registration time reduced from over thirty seconds to less than five seconds. The new methods are therefore suitable for image guidance in minimally invasive therapy. Significantly, these parallel global optimization approaches are also applicable to bioinformatics, computational chemistry, physics, econometrics, and in determining the appropriate parameters for mathematical models that fit experimental data.
4:25 PM - Coffee Break
Beverly Brechner (University of Florida)
Title: "Mathematics and Medicine"
This talk will discuss my work with the Brain Institute at the University of Florida. It will include two projects on which I have worked with them. The first project was on "Stereotactic Radiosurgery for Brain Tumors". The surgery is noninvasive, using radiation to kill certain kinds of brain tumors. Our problem was to automate the process. The (necessarily, semi-)automation was accomplished by my Ph. D. student, Taeil Yi. I will provide a complete overview of the process. The second project was to help automate (this time, invasive) surgery for Parkinson's disease. I worked closely with one of the Brain Institute's students, Atchar Sudhyadhom. I will discuss the surgery, the problems we encountered, and why automation may not be accomplishable at this time.