Detection of Brain Tumor using Magnetic Resonance images

Brain tumor is a fatal neuro-degenerative condition having only 19% of survival rate even after a significant diagnosis. Although diagnosing such a critical disorder is quite risky, but Magnetic Resonance Images (MRI) are fortunately using in most clinical diagnose centers for the identification, classification and segmentation of brain tumors in patients.

Brain tumor is one of the fatal neuro-degenerative disorders caused by unnatural proliferation of neurons. This disease is classified into several categories according to their structure and severity. This tumor is an unnatural cell growth in cerebrum and such brain disorder whose reason is still unidentified. Classification of Brain tumor is a process of removing the tumor from brain tissues involving clinical treatments like surgery, chemo-therapy and radio-therapy. However, such clinical diagnosis process is quite difficult due to the irregular form and undifferentiated location of tumors in deep brain regions. Although, many medical imaging modalities have been emerged to classify the characteristics of tumors; Magnetic Resonance Images (MRIs) are immensely used in medical image analysis for being non-invasive as it displays high resolution spatial images. It is not only advantageous in both visual and computer-aided image analysis, but also but turns out as one of the most upgraded technique for brain tumor detection, classification and diagnosis. According to their structure and rate of severity, tumors are primary and secondary. The primary stage tumor initiates either inside the brain or around the brain but don’t infect other parts of the body. Primary stage tumor can be reclassified into Malignant and Benign. The tumors at Benign is slow progressive having non-cancerous symptoms and needs no surgery where as Malignant has aggressive cancerous symptoms causing sever effect on brain functionality and can be diagnosed only though surgery. On the other hand, secondary stage tumor starts in any organ of the body and spread to the brain to blood streams.

This metastatic form of tumor is cancerous and life-threatening. Despite of that, there are more 120 numbers of tumors according to brain cell grading, incubation period, aggressiveness and treatment planning. There are many ways available for brain imaging such as ultrasound, CT scan, PET scan, MRI scan; but analysis of MRI data are more appropriate for visualizing brain lesions and tumors in a 3-dimensional axes. These data are enriched with high resolution enhanced spatial data depicting brain abnormalities which helps in medical imaging, classifying tumors according to their tissue characterization, identification of viable tumors etc. While analyzing MRI data, tumors are incorporated with different location, shape, depth of infection which is evident for effective classification. However, when a large amount of MRI data are scrutinizing for visual analysis; there is a huge chance of data misclassification. In aid to that, reduction in sensitivity and specificity during analysis of confusion matrix of a MRI dataset leads to incorrect classification. Recently, various machine learning and neural network methods are implicated in these dataset for classification, segmentation and detection process. But, compared to ML methods, NN techniques are mostly used with higher classification accuracy. Hence, most of the ML algorithms predominantly involve a neural network model to train and classify the images with significant accuracy.

References:
1. Padhy, Aditya & Mishra, Sandeep. (2022). Efficient Classification of Brain Tumor from Magnetic
Resonance Images with Improved Salp Swarm Algorithm and Convolutional Neural Network classification.
NeuroQuantology. 20. 774-782. 10.14704/nq.2022.20.3.NQ22364.
2. M. Tubishat, S. Ja’afar, M. Alswaitti, S. Mirjalili, N. Idris, M. A. Ismail, et al., “Dynamic salp swarm
algorithm for feature selection,” Expert Systems with Applications, vol. 164, p. 113873, 2021