Brain magnetic resonance images tumor detection based on lateral ventricles deformation analysis

This research is focused on developinga new approach to detecting, predicting and segmenting brain tumors in magnetic resonance images (MRI). The work is motivated by potential applications in assessing the shapes deformation of brain lateral ventricles and their correlation with tumor existence, examining treatment responses, enhancing computer-assisted diagnosis and surgery, planning radiation therapy, and constructing tumor growth models. Further implementation of this work may create the dynamic brain atlas in a three-dimensional view which is associated with brain tumor and the lateral ventricles and more advanced level, i.e., white matter (WM) and gray matters (GM). The presented framework forms brain MRI pre-processing, brain lateral ventricles segmentation, lateral ventricles deformation analysis and finally image classification. The key advantage of this framework is that the analysis on lateral ventricles shape deformations caused by compression from tumors assists in minimizing the tumor detection and segmentation complexity and workload, and in the mean time the estimation of brain lateral ventricles shape deformation is served as an additional probability factor or property for classification methods, therefore leads to a more accurate tumor locating and segmentation.

To make the above mentioned framework available, major activities of this study conducted are:

• A comprehensive pre-processing: applying a series of processing of whole brain extraction, intensity normalization and brain reorienting.
• Modified brain ventricular apartments segmentation methods for both template (normal) and diagnostic MR images: a new automatic Fuzzy C-Means (FCM) clustering scheme based on the modified FCM algorithms which are suitable for brain MRI ventricles segmentation and extraction.
• Ventricle deformation analysis: methods of using Thin Plate Splines (TPS) have been used to calculate the ventricle deformation in a quantitative manner. The TPS displacement data from the deformation analysis are retrieved to be served as an additional feature used for the further step of brain tumor classification.
• Classification for tumor detection and segmentation: Rough Set method has been applied to perform the classification and Support Vector Machines (SVM) will be used as another classification method.

Publications

• Xiao K, Ho S.H., Bargiela A., Automatic brain MRI segmentation scheme based on feature weighting factors selection, Intern. J. of Computational Intelligence in Bioinformatics and System Biology , (accepted for publication), 2009.
• Xiao, K., Ho, S. H., Bargiela, A. (2009) Brain Tumor Segmentation: Using Feature of Lateral Deformation. International Journal of Computer Assisted Radiology and Surgery, In Progress.
• Xiao, K., Ho, S. H., Bargiela, A (2008). Brain MRI tumor segmentation with the assistance of lateral ventricular deformation estimation. Journal of Australasian College of Physical Scientists and Engineers in Medicine, Accepted.
• Xiao, K., Ho, S. H., Hassanien, A. E. (2008). Automatic unsupervised segmentation methods for MRI Based on modified fuzzy c-means. Fundamenta Informaticae, Vol. 87(3-4):465-481.
• Xiao K., Ho S.H., Hassanien A.E., “Brain Magnetic Resonance Image Lateral Ventricles Deformation Analysis and Tumor Prediction”, Malaysian Journal of Computer Science, Vol. 20(2), 2007, pp. 115-132
• Xiao, K., Ho, S. H. "Brain Lateral Ventricles Shape and Tumor - Brain Tumor Prediction and Segmentation with the Assistance of Deformation-Based Morphometry Approach” book chapter in “Foundation on Computational Intelligence”, accepted by Series “Studies in Computational Intelligence”, Springer Verlag, Germany, 2008.
• El-dahshan E., Redi A., Hassanien A.E., Xiao K., “Accurate Detection of Prostate Boundary in Ultrasound Images Using Biologically-Inspired Spiking Neural Network”, in Proceedings of 2007 International Symposium on Intelligent Signal Processing and Communication Systems, Xiamen, China, Nov.28-Dec.1, 2007, pp. 333-33
• Xiao, K., Ho, S. H., Hassanien, A. E., Nguyen, V. D., Salih, Q. (2007). Fuzzy c-Means clustering with adjustable feature weighting distribution for brain MRI ventricles segmentation’, In: Proceedings of the ninth IASTED international conference on signal and image Processing. Pages: 483-489.
• Xiao, K., Ho, S. H., and Salih, Q. (2007). A Study: segmentation of lateral ventricles in brain MRI using fuzzy c-means clustering with Gaussian smoothing. In: Proceeding of the Joint Rough Set Symposium JRS2007, Lecture Notes in Computer Science, 4482:161-170.

To download a PDF file of the poster please click on the image