A deep learning approach for brain tumor classification using MRI images. (July 2022)

• Record Type: Journal Article
• Authors: Aamir, Muhammad; Rahman, Ziaur; Dayo, Zaheer Ahmed; Abro, Waheed Ahmed; Uddin, M. Irfan; Khan, Inayat; Imran, Ali Shariq; Ali, Zafar; Ishfaq, Muhammad; Guan, Yurong; Hu, Zhihua
• Is Part Of: Computers & electrical engineering. Volume 101(2022)
• Publication Date: 2022-07
• Subjects: Healthcare -- Deep learning features -- Feature fusion -- Illumination boost -- Non-linear stretching -- Localization -- Refinement -- MRI -- Brain tumor classification -- CAD; Computer engineering -- Periodicals; Electrical engineering -- Periodicals; Electrical engineering -- Data processing -- Periodicals; Ordinateurs -- Conception et construction -- Périodiques; Électrotechnique -- Périodiques; Électrotechnique -- Informatique -- Périodiques; Computer engineering; Electrical engineering; Electrical engineering -- Data processing; Periodicals; Electronic journals; 621.302854
• DOI: 10.1016/j.compeleceng.2022.108105
• Languages: English
• ISSNs: 0045-7906

Abstract

Highlights: An improved automated method for classifying brain tumors is proposed. An effective way to enhance the visual quality of MRI images is utilized. A system for locating objects (tumors) generates fewer but better proposals were developed. The hybrid feature vector is generated to improve the overall classification performance. The impact of overfitting on classification performance was explored. Comparisons with existing methodologies demonstrated that this strategy had greater classification precision. Abstract: Brain tumors can be fatal if not detected early enough. Manually diagnosing brain tumors requires the radiologist's experience and expertise, which may not always be available. Furthermore, manual processes are inefficient, prone to errors, and time-taking. Therefore, an effective solution is required to ensure an accurate diagnosis. To this end, we propose an automated technique for detecting brain tumors using magnetic resonance imaging (MRI). First, brain MRI images are pre-processed to enhance visual quality. Second, we apply two different pre-trained deep learning models to extract powerful features from images. The resulting feature vectors are then combined to form a hybrid feature vector using the partial least squares (PLS) method. Third, the top tumor locations are revealed via agglomerative clustering. Finally, these proposals are aligned to a predetermined size and then relayed to the head network for classification. Compared to existing approaches, the proposed method achieved a classification accuracy of 98.95%. Graphical Abstract: Image, graphical abstract