Student : Adrian – Alin BarglazanPaper :”Wall position and thickness estimation from sequences of images” - Dias, J.M.B.; Leitao, J.M.N.;
WALL POSITION AND THICKNESS ESTIMATION FROM SEQUENCES OF IMAGES
ECHOCARDIOGRAPHY - IMPORTANCE Ventricular contours Volume of chambers Thickness of myocardium Ventricular mass 3D reconstruction/modeling through
cardiac cycle
ECHOCARDIOGRAPHY
Advantages : Noninvasive agent Low cost Portability Real-time processing Direct 3D acquisition
ECHOCARDIOGRAPHY – MAIN DEGRADATION MECHANISMS
Side lobes Blur Poor Contrast Artifacts Speckle noise
PREVIOUS WORK
“Semiautomatic border tracking of cine echocardiogram ventricular images” – D.Adam, H. Harauveni, S. Sideman – 1987 :
Non linear median filter(9X9) of whole images. Location-dependent contrast stretching Tracks the movement of predetermined points which are manually defined
on the 2 myocardial border “Detecting left ventricular endocardial and epicardial
boundaries by two-dimensional” – C. Chu, E. Delp Edge detector – 41x41 Gaussian filter folowed by a Laplacian operator
The noise effects(speckle effect in principal) make conventional techniques based on edge enhancement inappropriate – gradient threshold, Laplace
PREVIOUS WORK
“Automated extraction of serial myocardial borders from an M-mode echocardiograms” – M. Unser, G. Pelle, P. Brun, M. Eden – 1989 :
Used suitable matched filters “Automatic ventricular cavity boundary detection from
sequential ultrasound images using simulated annealing “ - D. Adam
Proposed a fully automatic boundary detection from sequential images using simulated annealing .
PROPOSED APROACH
Image characterization – given a tissue, image is considered pixel wise independent .
Heart morphology – for example if we scan from inside to outside the values of the pixel should have a rectangular shape.
Contour model – contour sequences are assumed 2 dimensional Markov processes. Each random variable has a spatial index and a temporal index
Bayesian formulation and MAP IMDP – iterative multigrid dynamic programming –
to solve the problem of optimization
PROBABILISTIC MODEL OF ENDOCARDIAL AND EPICARDIAL CONTOURS
Represent the contour
Polar coordinates Heart contour Reflectivity
PROBABILISTIC MODEL OF ENDOCARDIAL AND EPICARDIAL CONTOURS
Echo along a radial scan-line from the heart center towards lung tissue.
THE MAIN ALGORITHM
RESULTS
RESULTS