Automatic Brain Tumor Segmentation Using Tissue Diffisivity Characteristics

Abstract

Water diffusion measurements have been shown to be sensitive to tissue cellular size, extra cellular volume, and membrane permeability. Therefore, diffusion tensor imaging (DTI) by MRI can be used to characterize highly cellular regions of tumors versus acellular regions, distinguishing cystic regions from solid regions. An automatic segmentation method is proposed in this paper based on a multi-phase clustering algorithm to segment the brain tumors in a feature space extracted from DTI images. The algorithm is applied on images of a total of 20 patients with 4 different types of tumors. The tumor region segmentation was 92% accurate based on the segmentation results using anatomical images and 100% accurate based on biopsy results. In general, the segmentation results obtained by the proposed method revealed a strong agreement with the biopsy results and anatomical images, providing support for the accuracy and robustness of the proposed feature space and the segmentation procedure.

The Automatic Brain Tumor Segmentation Based on MRI Using Optimal Morphology Thresholding Methods

“…Magnetic resonance imaging (MRI) is a tool with high resolutions and commonly used for identifying the abnormalities of tissues more efficiently than other imaging methods, for example, diagnosing, making treatment plans for and following up the brain tumors under the variety of imaging methods, selecting the thickness of slices and the features of MRI scanners [13]. Although the efficient segmentation and the identification of the brain tumors were not as important as treatments and plans; the segmentation and the identification were useful for evaluating and following up treatments, improving the efficiency of the identification with the color changes as well as the different brightness, contrast, sizes and locations of brainstems.…”

Automated Localization of Brain Tumors in MRI Using Potential-K-Means Clustering Algorithm

“…It tends to perform better in narrow domains, e.g., medical images, where visual variability is limited. Most previous work on automated brain tumor segmentation is based on machine learning, both supervised [1][2][3][4][5][6][7][8] and unsupervised [9][10][11][12][13][14][15][16][17]. Among the supervised methods, approaches include using Support Vector machines [3,4], Bayesian classifier [5], fractal features [6], outlier detection [7], Markov Random Fields [8].…”

Segmentação de tumores de encéfalo em imagens por ressonância magnética baseada em informações texturais.

“…A etapa de classificação normalmente nãoé perfeita, uma vez que o classificador pode atribuir erroneamente elementos do conjunto de testesà classe incorreta, gerando pontos ou pequenas regiões conexas na segmentação final. A forma mais simples de tratar tal problemaé utilizando-se operações morfológicas como dilatação e erosão (ZHOU et al, 2005), apesar de existirem métodos mais complexos como o uso de filtros de mediana (SCHMIDT, 2005) ou limiarização local seguida de crescimento de região (SHAHMORAD et al, 2007). Por serem rápidas e eficientes, as operações morfológicas são adotadas na etapa de pós-processamento deste trabalho, para refinar a segmentação obtida.…”

“…Já no regime de teste inter-pacientes (foram usados dados de diferentes pacientes para treinamento e teste) obteve-se acurácia de 98% para edema, e 73% para tumor realçável. Já em(SHAHMORAD et al, 2007), os autores propuseram um método para segmentação automática de regiões de tumor em imagens também do tipo DTI. Nele, foi empregado um algoritmo de k-médias, de aprendizado não-supervisionado, para realizacão da segmentação automática.…”