edge-basedalgorithmofpeakpositionsearchingforpet-csus.pptVIP

Department of Computer Science EDGE-BASED PEAK POSITION SEARCH ALGORITHM FOR PET DETECTORS Presenter: Kun Di Advisor: Dr. Chung-E Wang Dr. Ted Krovetz Department of Computer ScienceCalifornia State University, Sacramento November 22, 2010 * Department of Computer Science * Agenda Motivation Background Methodology Edge Detection Gradient Information Analysis Peaks Detection Results Summary and Future Work * Department of Computer Science * Motivation The positron emission tomography (PET) scanner Gamma Rays Array of Scintillation Crystals Photomultiplier Tubes (PMTs), Avalanche Photo Diodes (APDs) Coincidence processing Calibration: Position Profile (Peak Positions) Automatic peak detection is desirable because manually selecting each peak is time-consuming, especially for large crystal arrays (e.g. 20x20). Pictures (except the right-bottom one) are from /wiki/Positron_emission_tomography * Department of Computer Science * Background Previous Effort Neural network based algorithm[2] Non-rigid registration to a Fourier-based template [3] PCA-based algorithm [4] Morphological-based algorithm [5][6] Main obstacles Distortion Peak fusion. * Department of Computer Science * Edge Detection Canny Edge - Continuous Edge Gaussian Smooth Gradient (magnitude) Non-Maximum Suppression High and Low Threshold Trace Edge Gradient * Department of Computer Science * Gradient Information Magnitude Direction Special Points Top, Bottom, Left, right Break Points Horizontal, Vertical Gradient Information Analysis * Department of Computer Science * Peaks Detection Algorithm Raw image input; Generate Gaussian smoothed image; Edge detection according to gradient magnitude; Get interested region according to the edges and gradient directions of each point on each edge; Identify peaks according to the number of crystals and hit-sum of each interested region; Manually correct the peaks if necessary; Save the peaks; * Department of Computer Science * Tools Implementatio