Computer-Aided Diagnosis System for Automated Detection of Mri Brain Tumors
Umar S. Alqasemi1, Sultan A. Almutawa2, Shadi M. Obaid3
1Umar S. Alqasemi, Department Electrical and Computer Engineering, King Abdulaziz University, Jeddah, Makkah, Saudi Arabia.
2Sultan A. Almutawa, Department Electrical and Computer Engineering, King Abdulaziz University, Jeddah, Makkah, Saudi Arabia.
3Shadi M. Obaid, Department Electrical and Computer Engineering, King Abdulaziz University, Jeddah, Makkah, Saudi Arabia.
Manuscript received on 02 August 2023 | Revised Manuscript received on 19 January 2024 | Manuscript Accepted on 15 February 2024 | Manuscript published on 28 February 2024 | PP: 40-38 | Volume-13 Issue-3, February 2024 | Retrieval Number: 100.1/ijeat.C436013030224 | DOI: 10.35940/ijeat.C4360.13030224
Open Access | Editorial and Publishing Policies | Cite | Zenodo | OJS | Indexing and Abstracting
© The Authors. Blue Eyes Intelligence Engineering and Sciences Publication (BEIESP). This is an open access article under the CC-BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/)
Abstract: The detection and classification of brain tumours manually or traditionally is an area that could be improved by having an automated detection and classification system for brain tumours. In this paper, an enhanced Computer-Aided Diagnosis CAD software system is introduced for brain tumour detection and classification. A total of 229 brain MRI images were taken as the dataset for this research; these dataset images include 105 normal brain MRI images and 124 abnormal brain MRI images. The proposed CAD system is specialised for the detection and classification of Meningioma brain tumours. The technique can be generalised and implemented for Glioma and Pituitary brain tumours as well. The entire system was implemented using MATLAB software. We began by cropping the region of interest (ROI) from the dataset images. Then, feature extraction was implemented using first-order statistical features, as well as the use of wavelet filters in combination with these features. The t-test is used to exclude features of no statistical significance (p-value < 0.05). After that, different types of classifiers were used to separate the standard set from the abnormal one. Note that we employed an iterative approach, changing features through multiple runs, until we achieved the best performance. The best accuracy results were obtained with the SVM-Kernel Function (Linear), KNN-1, KNN3, and KNN-5 classifiers. Note also that we used convolutional neural networks (CNNs) from the Deep Learning toolbox of MATLAB as a control method to compare, where the images were fed directly into the CNN. The results were evaluated using performance assessment techniques, including Sensitivity, Specificity, Positive Predictive Value (PPV), Negative Predictive Value (NPV), Accuracy, Error Rate, and Area Under the Curve (AUC) of the Receiver Operating Characteristic (ROC). With the SVM classifier, the best accuracy results were 91%, followed by the CNN classifier at 82%, and the KNN classifier at 77%. Furthermore, it was beneficial to find such feature extraction techniques that yielded acceptable accuracy results with three different classifiers; this was the case twice, as mentioned in the study. All proposed CAD system areas were developed and implemented using MATLAB software.
Keywords: CAD System, MATLAB, Automated Detection and Classification, MRI Brain Tumors, SVM Classifier, KNN Classifier, CNN, DCNN
Scope of the Article: Classification