An automated brain tumor classification in MR images using an enhanced convolutional neural network

• Published in: International journal of information technology (Singapore. Online) Vol. 15; no. 2; pp. 665 - 674
• Main Authors: Singh, Ravendra, Agarwal, Bharat Bhushan
• Format: Journal Article
• Language: English
• Published: Singapore Springer Nature Singapore 01.02.2023; Springer Nature B.V
• Subjects: Accuracy; Artificial Intelligence; Artificial neural networks; Automation; Brain; Brain cancer; Brain research; Classification; Computer Imaging; Computer Science; Decision support systems; Glioma; Image Processing and Computer Vision; Image segmentation; Machine Learning; Magnetic resonance imaging; Medical imaging; Neural networks; Original Research; Pattern Recognition and Graphics; Software Engineering; Tumors; Vision; Brain tumor detection systems; Segmentation; MRI; Convolutional neural network; Computer-aided diagnosis; Classification
• ISSN: 2511-2104; 2511-2112
• DOI: 10.1007/s41870-022-01095-5

MRI is a non-invasive imaging tool, accurate classification of brain tumours from MRI images is a highly specialized area of a medical study. Classification of brain tumours is a method for identifying and automatically labeling malignant brain tissues based on the types of tumours present. Gliomas can be diagnosed and treated using magnetic resonance imaging (MRI) in clinical practice. The ability to appropriately classify a brain tumour from MRI images is critical to clinical diagnosis and therapy planning. However, due to the vast volume of data generated by MRI, manual classification is not possible promptly. For this reason, it is necessary to use automated methods for classification and segmentation. However, MRI image segmentation is complicated by the wide range of spatial and structural variations across brain tumours. For the categorization of three distinct forms of brain tumors, we’ve developed a novel convolutional neural network (CNN) architecture. MRI images with contrast-enhanced T1 images were used to show that the new network was simpler than previous networks. The network’s performance was evaluated using two ten-fold cross-validation procedures and two datasets. As part of the subject-wise cross-validation procedure, an improved image database was employed to test the network’s generalizability. This method of ten-fold cross-validation data set has an accuracy rating of 92.50% when applied to record-wise cross-validation. A new generalization potential and rapid execution of the newly proposed CNN architecture in medical diagnostics imply that it may be a useful decision-support tool for radiologists working in the field of medical diagnostics.