Submit Manuscript  

Article Details


Segmentation of Brain Magnetic Resonance Images using Deep Learning Classification and Multi-modal Composition

Author(s):

Kala R and Deepa P*  

Abstract:


Background: Brain tumor detection and accurate identification of its severity is a challenging task for radiologists as reliable brain tumor segmentation is a critical component in cancer diagnosis, treatment planning, and treatment outcome evaluation.

Method: To build upon successful deep learning techniques, a brain tumor segmentation method is developed by using the deep convolutional neural networks by taking input MR images with different modalities such as Flair, T1, T1C and T2. The images from the different modalities are used in proportion to the information content in the particular modality. The weights for the different modalities are calculated blockwise and the standard deviation of the block is taken as a proxy for the information content of the block.. Then the convolution is performed between the input image of the flair, T1, T1C and T2 MR images and corresponding to the weight of the flair, T1, T1C and T2 MR images. The convolution are summed between the corresponding input and weight of the MR images to obtain a new composite image which is given as a input image to the deep convolutional neural network to obtain segmentation results with better appearance and spatial consistency. The analysis of the proposed method shows that the discriminatory information from the different modalities are effective combined to increase the overall accuracy of segmentation.

Results: The proposed method was validated in the Brain Tumor Segmentation Challenge 2013 database (BRATS 2013), obtaining the complete, core and enhancing regions in Dice Similarity Coefficient and Jaccard similarity index metric for the Challenge, Leaderboard and Synthetic data set. To evaluate the classification rates, the metrics such as accuracy, precision, sensitivity, specificity, under segmentation, incorrect segmentation and over segmentation also evaluated and compared with the existing methods. Experimental results exhibit a higher degree of segmentation accuracy compared to existing methods.

Keywords:

Convolutional neural networks, Brain tumour, Deep learning, Magnetic resonance image, Classification, Modalities.

Affiliation:

Electronics and Communication Engineering, Government college of technology, Coimbatore, Tamilnadu 641 013, Electronics and Communication Engineering, Government college of technology, Coimbatore, Tamilnadu 641 013



Full Text Inquiry