Brain Tumor Detection & Classification Using Extreme Machine Learning and Implementation through Embedded System.

Abstract

The field of medical imaging is gaining importance with an increase in the demand for automated, reliable, fast and efficient diagnosis which can provide insight to the image better than human eyes. Brain tumor is the second leading cause for cancer-related deaths in men in age 20 to 39 and fifth leading cause cancer among women in same age group. Diagnosis of brain tumor is a very important part in its treatment. A prime reason behind an increase in the number of cancer patients worldwide is the ignorance towards treatment of a tumor in its early stages. The manual detection and classification of the tumor becomes a rigorous and hectic task for the radiologists. The segmentation, detection, and extraction of infected tumor area from magnetic resonance (MR) images are a primary concern, but a tedious and time taking task performed by radiologists or clinical experts, and their accuracy depends on their experience only.The FCM (Fuzzy C Means) based segmentation techniques such as spatial information(FCM_S), FCM_S1 and FCM_S2, enhanced FCM algorithm (EnFCM), FGFCM (Fast generalized FCM algorithm), fuzzy local information c-means clustering algorithm (FLICM) has ability to obtain texture and background information but failed to remove in the case of complex rician noisy images. FLICM, improves the segmentation process, but fails to remove Gaussian noise beyond 30%. To improve the performance and reduce the complexity involved in the medical image segmentation process, a novel fast and robost FCM (FRFCM) image segmentation has been investigated in this research work. Furthermore, to improve the accuracy and quality rate of the neural network-based classifier, relevant features are extracted from each segmented tissue and aligned as input to the classifiers for automatic detection and classification of brain tumors. The simulation results of proposed technique have been evaluated and validated for performance and quality analysis on magnetic resonance brain images, based on feature coefficient.The novel Fast and robust Fuzzy C Means (FRFCM) segmentation technique has been proposed for detection of brain tumor from MR (Magnetic Resonance) images that can inform the radiologist and doctor about the details of brain tumor. This segmentation technique include noise removal and sharpening of the MR image along with basic morphological functions, erosion and dilation, to obtain the background. The MR (Magnetic Resonance) Images featuresivhave been extracted through a popular Gray Level Co-occurrence Matrix (GLCM) feature extraction technique. The extracted features are applied to the, proposed PSO+ELM (Particle Swarm Optimization-Extreme Learning Machine) for classification of the type of malignant (cancerous) and benign (non-cancerous) brain tumors for visual localization. The weights of the proposed novel multi class extreme learning machine classifier model has been updated by the PSO (Particle Swarm Optimization) algorithm to increase the performance of the classifiers. Further the classification results have been compared with the existing support vector machine and relevance vector machine model sparse Bayesian extreme learning machine. The proposed PSO+ELM model obtained an accuracy of 99.41% which is higher than the other said models. To validate the robust ness and rician noise removal capability of the proposed FRFCM segmentation two quality indexes are considered as Structural Similarity (SSIM) index and the Quality Index based on Local Variance (QILV). However, SSIM is more sensitive to the noise level in the image and the QILV to blurring of the edges. In addition to the both, the PSNR (Peak signal to noise ratio) is also calculated. It is also observed that the quality measure PSNR value is 37.32 dB and SSIM as 0.9219 dB. The higher value of PSNR indicate better signal-to noise ratio in the extracted image. Also, the larger value of SSIM indicates the noise reduction in the extracted image.The automatic segmentation, feature extraction and classification has also been presented through GUI (Graphical User Interface) using MATLAB2018b software. Further, the detection and classification has been implemented through the embedded system platform by employing Raspberry PI3B+ along with python which may be the prototype product outcome of the research work. It will help the medical staff, particularly for the radiologist and doctor to understand the seriousness of the tumor.