Multimodal Visualization for neurosurgical planning CMPS 261 June 8th 2010 Uliana Popov Motivation Each year more than 200,000 people in the United States are diagnosed with a primary or metastatic brain tumor. Brain cancer remains one of the most incurable forms of cancer, with an average survival period of one to two years. The chances of surviving for a person with a brain tumor greatly depends on all of the following: type of tumor What is a brain tumor? A group (mass) of abnormal cells that starts in the brain There are over 120 different types of brain tumors, which makes effective treatment complicated. Diagnosis of brain tumors Diagnostic tools include: patient history, a brain scan, CT scan, MRI. MRI provides a much greater contrast between the different soft tissues of the body than computed tomography (CT) does The first MR image was published in 1973 MRI How does it work? The body is largely composed of water molecules. Each water molecule has two hydrogen nuclei or protons A powerful magnetic field causes the magnetic moments of some of these protons to align with the direction of the field. The protons in different tissues return to their equilibrium state at different rates MRI sequences echo time - TE repetition time – TR T1 T1-weighted scans use a gradient echo (GRE) sequence, with short Te and short Tr This scan runs very fast allowing the → easy too collect high resolution 3D datasets. T2 T1-weighted scans provide good gray matter/white matter contrast. MRI sequences (cont) Diffusion MRI Diffusion MRI measures the diffusion of water molecules in biological tissues. If molecules in a particular voxel diffuse principally in one direction → the majority of the fibers in this area are going parallel to that direction. Fluid Attenuated Inversion Recovery (FLAIR) Inversion-recovery pulse sequence used to null signal from fluids. Susceptibility weighted imaging (SWI) Produces an enhanced contrast magnitude image very sensitive to venous blood, hemorrhage and iron storage. Used to enhance the detection and diagnosis of tumors Goal Determine: Number of tumors Number of abnormal cells* Tumor margins *voxels Previous Work Probabilistic segmentation of brain tumors based on multi-modality magnetic resonance images. 2007 Acqire results using 21 patients. Test it on 22nd 2D 3D brain tumor segmentation in MRI using fuzzy classification, symmetry analysis and spatially constrained deformable models. 2008 My approach Combine together *voxel Apply symmetry criteria to filter out ”not interesting” regions Use multimodality to vote for each cell* Data IEEE contest organizers provided the data* MRI sequences Sequence name Dimensions Data type Voxel to World matrix * The dataset is courtesy of Prof. B. Terwey, Klinikum Mitte, Bremen, Germany. Data (cont) The process For each sequence in the data Convert to physical space Register Calculate Symmetry Calculate Gradient The process, step 1 Convert from computational space to physical The process, step 2 Registeration MedINRIA Algorithm: manual landmark based Non linear transformation The process, step 3 Find asymmetry : If a[i,j] ~= a[n-i, j] a[i,j] = 0; a[n-i,j] = 0; The process, step 4 Calculate gradient Data values are scalars Scalar' = Vector Calculate Magnitude of the vector Process (cont) Reference – healthy brain Generated by using BrainWeb* - Simulated MRI Volumes for Normal Brain *McConnell Brain Imaging Centre (BIC) of the Montreal Neurological Institute, McGill University. Process (cont) Calculate probability Voting function If Healthy cell gradient < X && tumor cell gradient > Y Higher probability that the cell is abnormal Treat noise clear skull boundaries manual registration Results Results (cont) Voxel size 0.924mm x 1.14169mm x 2.38699mm Total brain voxels 1,222,332 Tumor voxels 14,878 → 0.012171816 % Tools Algorithm implementation C++ using VTK libraries Viewers Paraview Registration MedINRIA Future Work Make a tool Add DTI processing Determine spatial relationship between tumor and WM fiber tracts