Brain tumors can be detected and classified using invasive medical procedures like biopsy. However, biopsy is not usually conducted before definitive brain surgery. With the drastic improvement in the field of computer science and artificial intelligence, algorithms can be developed in order to identify and perform the tumor diagnostics without any need for invasive measures. Convolutional neural network(CNN) is one such algorithm that has shown substantial scope in this very field. The performance of our algorithm on BraTS dataset which is publicly available MRI image dataset (n=3000) is promising in this domain. Our approach is compared with previous classical machine learning and deep learning efforts. The approach discussed in this paper remarkably obtained an accuracy of 0.9868 and outperformed previous methods.

A novel CNN architecture was presented in this study which can be used for automated brain tumor classification using MRIs. The classification was performed using an Axial MRI image database. As input, we used whole images, then preprocessed the images and went forth with identification and segmentation. Our proposed neural network has outperformed classical models [8], it is simpler than classical pre-trained CNNs and can be executed on modern conventional PCs. The main reason being, our proposed approach utilizes fewer resources for training and testing purpose. It is always beneficial for developing countries to adapt models that require fewer resources, approaches that use smaller networks are preferable for such countries [7]. As in our proposed approach, due to less resource utilization, it can be deployed on mobile platforms as well. The model trained and validated around 3000 images in batch sizes of 32. The loss obtained was 0.0373 and the accuracy obtained was 0.9868. The validation loss obtained was 0.2695 and the validation accuracy was 0.9450. In addition, the network has a very good execution speed of 0.1 seconds per image.

1.Bleyer, W. Archie. “Epidemiologic impact of children with brain tumors.” Child’s nervous system 15.11 (1999): 758-763. 2.Sharif, Muhammad Imran, et al. “A decision support system for multimodal brain tumor classification using deep learning.” Complex & Intelligent Systems 8.4 (2022): 3007-3020. 3.Menze, Bjoern H., et al. “The multimodal brain tumor image segmentation benchmark (BRATS).” IEEE transactions on medical imaging 34.10 (2014): 1993-2024. 4.Badža, Milica M., and Marko Č. Barjaktarović. “Classification of brain tumors from MRI images using a convolutional neural network.” Applied Sciences 10.6 (2020): 1999. 5.Ezhilarasi, R., and P. Varalakshmi. “Tumor detection in the brain using faster R-CNN.” 2018 2nd International Conference on I-SMAC (IoT in Social, Mobile, Analytics and Cloud)(I-SMAC) I-SMAC (IoT in Social, Mobile, Analytics and Cloud)(I-SMAC), 2018 2nd International Conference on. IEEE, 2018. 6.Abd-Ellah, Mahmoud Khaled, et al. “Two-phase multi-model automatic brain tumour diagnosis system from magnetic resonance images using convolutional neural networks.” EURASIP Journal on Image and Video Processing 2018.1 (2018): 1-10. 7.Akram, M. Usman, and Anam Usman. “Computer aided system for brain tumor detection and segmentation.” International conference on Computer networks and information technology. IEEE, 2011. 8.Díaz-Pernas, Francisco Javier, et al. “A deep learning approach for brain tumor classification and segmentation using a multiscale convolutional neural network.” Healthcare. Vol. 9. No. 2. MDPI, 2021.