Work

CNN-Based System for Automated Detection and Segmentation of Brain Tumors: A Performance Study

Year: 2024
Type: article
Abstract: Brain tumor classification using magnetic resonance imaging (MRI) is a vital step in assisting clinicians with early diagnosis and treatment planning. Manual examination of MRI scans is often time-consuming and prone to observer variability, which emphasizes the need for automated diagnostic methods. In this study, a convolutional neural network (CNN)-based framework was developed to perform multi-class classification of brain tumors, distinguishing between glioma, meningioma, pituitary tumor, and no tumor. A systematic preprocessing pipeline was applied, including grayscale conversion, resizing, normalization, and augmentation, to standardize the dataset and improve model robustness. Several baseline CNN models with varying complexity were designed and evaluated, and insights from these experiments guided the development of a deeper proposed CNN architecture. Performance evaluation incorporated accuracy, precision, recall, F1-score, and receiver operating characteristic (ROC) analysis, alongside visualization through confusion matrices. The findings highlight that the proposed CNN provides substantial improvements over baseline models and demonstrates strong capability in extracting discriminative features from MRI scans. Overall, the study confirms that CNN-based approaches hold significant promise for reliable and efficient brain tumor classification, offering a pathway toward clinical decision-support systems.
Author: Usha Barad
Language: English
Open Access status: diamond
OpenAlex ID: W7217788025
FWCI:
Cites: 53
Cited by:
Related to: 10