Medical Image Processing Studies: Current Research in Biomedicine Method of finding projects was to type in “image/signal processing” into several search engines and sift through the hits to find interesting/relevant topics
Research Univ. of Wisconsin – Madison Univ. of Minnesota ECG Compression using Signal Processing http://saigon.ece.wisc.edu/~waveweb/Research/ECG/page1.html Univ. of Minnesota Magnetic resonance imaging based on blood oxygen level dependent contrast New type of magnetic resonance imaging based on detection of image contrast determined by local varations in magnetic field Takes advantage of non-spatially uniform magnetization Animation of MRI: http://www.cmrr.drad.umn.edu/fMRImovie.html Univ. of Wisconsin: MATH ECG waveforms are signals measured from heart, providing essential info to cardiologist … normally 24-hr recording is desirable to detect abnormalities, compression is needed for storage… use percent-root-mean-square difference – crucial requirement is not to distort the diagnostic info used by the physician Their technique compresses the data by discarding relatively insignificant information Univ. of Minnesota: Used to generate functional images in human brain BOLD relies on magnetic field inhomogeneity generated when blood contains deoxyhemoglobin deoxyhemoglobin is generated when oxygen is used by cells in body, it is paramagnetic, therefore properties differ from the magnetic properties of tissue surrounding the blood vessels containing deoxyhemoglobin… this provides magnetic field gradients across the boundaries of the blood vessels For animation slides: Head of subject rendered as opaque volume, followed by the skull changing from translucent to transparent to show activation on cerebral cortex Using translucent opacity for the brain, while maintaining activation as opaque, 3d structure is illustrated Cerebrum fades out to leave only cerebellum, which is then replicated into 2 cerebellums, showing difference between subject completing difficult v. simple task
More Research… Stanford Univ. – Center for Advanced Medical Informatics Spinal Image Fusion Project Benefits: Improve diagnostic accuracy of spinal imaging techniques, improve outcome of existing treatments for patients with severe spinal disorders Semi-Automated Editing of Medical Images Benefits: Expedite editing process, save expensive radiologist time 3D Reconstruction of Blood Vessels from MR Angiograms Benefits: Classify MRA data into vessel & surrounding-tissue categories, then implementation of program to test segmentation methods http://camis.stanford.edu/projects/image_process.html Stanford U: Spinal Image Fusion Project: Enhance diagnostic capability of CT & MRI in area of spinal disorders Surgeons need to be able to visualize spatial relationships among various soft & hard tissue types associated with vertebrae next to disks & nerves Goal: generate automatically fused images of cervical spine by using retrospective registration technique Semi-automated Editing of Medical Images: 3d reconstruction procedure using edited images to generate clear displays of selected organs, minimizing editing time by the radiologist Algorithm: segmentation of all images, search procedure to find segments w/I sequence of images that belong to same organ 3D Reconstruction of Blood Vessels from MR Angiograms: Recent CT & MR angiography tech. Have provided 3d observations of vascular structures without arterial catheterization, but these images are often noisy … Developing methods that segment vessels using direction & magnitude of blood-flow velocity
Research cont… Yale University Univ. of Michigan Segmentation of Anatomical Structure Measurement of Cortex from 3D MR Images Boundary finding with Correspondence http://noodle.med.yale.edu/seg.html Univ. of Michigan Elasticity Imaging Can noninvasively detect kidney transplant rejection http://bul.eecs.umich.edu/research/index.html 1st project: quantitative analysis helps in understanding of brain anatomy & function, can be done by extracting structure & obtaining representation for various measurements Manual segmentation tedious and intensive, automatic segmentation is not developed fully yet Their solution: (taking into account the nearly constant thickness of the cortical mantle) evolve two embedded surfaces simultaneously, final representation and automatic segmentation can occur: Cortical surface area, surface curvature, and cortical thickness can be then evaluated Boundary finding: Based on prior knowledge and edge info. Of input image, used to find the boundary with subset points in correspondence w/ boundaries. Univ. of Michigan: New diagnostic modality to detect abnormalities in organs by using sensitive ultrasound speckle tracking procedures, controlled surface deformations and quantitative reconstruction algorithms – provide info. About mechanical properties of internal organs, such as kidney elasticity which can indicate renal damage - Can noninvasively detect kidney transplant rejection
More research… University of Alberta, CA Medical Imaging Group Acquiring & manipulation of image data for rehabilitation of scoliosis patients To see an example of the imaging: http://nyquist.ee.ualberta.ca/medimg/medimg.html IEEE Transactions on Medical Imaging Signal Processing Society Engineering in Medicine & Biology Society http://www.ieee-tmi.org/ This group works mainly at the Glenrose Rehabilitation Hospital in N. Edmonton, task is to acquire & manipulate image data for the rehabilitation of scoliosis patients
Programs in Image Processing: Magnetic Resonance User Interface Matlab based Graphical User Interface software Uses time domain signals, then FT for interpreting data and quantitation results http://www.mrui.uab.es/mrui/mruiAbout.html NASA’s Langley Research Center: Retinex Image Processing http://dragon.larc.nasa.gov/retinex/medical/medical1.html Graphical interface for the processing of in vivo MR data in time domain – eliminates limitations associated with Fourier Transformation methodologies Frequency selective filtering of signals, correction of artifacts, and quantitation by fitting peaks in spectrum
More programs… TRACEE Factor Analysis of Dynamic Structures Unix-based image processing environment Includes visualization & analysis tools http://physig.ph.kcl.ac.uk/tracee/overview.html Factor Analysis of Dynamic Structures Benefits: Technicians spared time drawing regions of interest around organs; computational speed, increased accuracy (speculatory) http://www.ph.kcl.ac.uk/ipg/fads.html Fully automated method of identifying and extracting overlaying structure in medical image sequences, mostly used for time series Uses temporal info to discriminate between structure which correspond to organ or other compartment in body Technique: Performing principal components analysis on pop. Of time-activity curve obtained from each pixel in the image… mathematical transformation
Interesting websites Westminster Dept. of Electrical Systems CMSA Filter Designer - http://www.cmsa.westminster.ac.uk/filter_design.html Signal Processing Information Database www.hr/josip/DSP/sigproc.html News of Diagnostic Imaging www.dimag.com Center for Imaging Science http://cis.jhu.edu/main.html Choose a cut-off freq, and desired filter, and it designs digital filter & draws the frequency response
More cool stuff… Medical Image Analysis Online www3.oup.co.uk/mediaj/ Bio-Imaging Research, Inc. Information about CT scans, ultrasound imaging, MRIs, and more http://www.bio-imaging.com/
Test Images Penn State University’s Electrical Engineering Dept. Multidimensional Image Processing http://cobb.ece.psu.edu/ Computer Vision Test Images http://www.cs.cmu.edu/~cil/v-images.html