Related Experiment Video For brain tumor
Updated: Nov 19, 2025
Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography
Published on: November 30, 2022
An efficient approach to diagnose brain tumors through deep CNN
Bakhtyar Ahmed Mohammed1, Muzhir Shaban Al-Ani2
1University of Human Development, College of Science and Technology, Department of Computer Science, Sulaymaniyah, KRG, Iraq and University of Sulaimani, College of Science, Department of Computer, Sulaymaniyah, KRG, Iraq.
Background And Objective:
Brain tumors are among the most common complications with debilitating or even death potential. Timely detection of brain tumors particularly at an early stage can lead to successful treatment of the patients. In this regard, numerous diagnosis methods have been proposed, among which deep convolutional neural networks (deep CNN) method based on brain MRI images has drawn huge attention. The present study was aimed at proposing a deep CNN-based systematic approach to diagnose brain tumors and evaluating its accuracy, sensitivity, and error rates.
Materials And Methods:
The present study was carried out on 1258 MRI images of 60 patients with three classes of brain tumors and a class of normal brain obtained from Radiopedia database recorded from 2015 to 2020 to make the dataset. The dataset distributed into 70% for training set, 20% for test set, and 10% for validation set. Deep Convolutional neural networks (deep CNN) method was used for feature learning of the dataset images which rely on training set. The processes were carried out through MATLAB software. For this purpose, the images were processed based on four classes, including ependymoma, meningioma, medulloblastoma, and normal brain.
Results:
The results of the study indicated that the proposed deep CNN-based approach had an accuracy level of 96%. It was also observed that the feature learning accuracy of the proposed approach was 47.02% in case of using 1 epoch, which increased to 96% when the number of epochs rose to 15. The sensitivity of the approach also had a direct relationship with the number of epochs, such that it increased from 47.02 to 96% in cases of having 1 and 15 epochs, respectively. It was also seen that epoch number had a reverse relationship with error rate which decreased from 52.98 to 4% once the number of the epochs increased from 1 to 15. After that the system tested on 25 new MRI images of each of the classes with accuracy 96% according to the confusion matrix.
Conclusion:
Using deep CNN for feature learning, extraction, and classification based on MRI images is an efficient method with an accuracy rate of 96% in case of using 15 epochs. It exhibited the factors which cause increase accuracy of the work.
