BRAIN TUMER DETECTION SYSTERM USING DEEP LEARNING

This research proposes an automatic, intelligent, hybrid system termed Auto Contrast Enhancer, Tumor Detector and Classifier for detecting and classifying brain tumors. The system utilizes the potential of transfer learning to obtain the optimal performance of the deep learning-based system even when a small dataset is available for training. In the first phase, the system employs ODTWCHE for pre- processing low-contrast MRI images, followed by brain tumor detection. In the second phase, the system utilizes the Inception V3 pre-trained model to automatically extract all high and low-level features from brain MRI images for brain tumor classification. The suggested system is efficient in accurately identifying and classifying the tumor because it integrates contrast enhancement techniques with deep transfer learning for brain tumor diagnosis and classification. It is a robust structure for automated identification and classification of brain tumors. The pro- posed system was tested on the openly accessible Fig share database [13] and achieved the highest accuracy of 98.89 % compared to existing systems in the literature [11–14], such as AlexNet [11], VGG-16 [12], VGG-19 [12], Dense Net [14], and ResNet-50 [15]. MRI images of varying contrasts were further used to assess the proposed system’s performance. On the basis of simulation results, the authors claim that the performance of the system does not degrade with the change in contrast and brightness of the MRI image, proving the system’s robust- ness. The Proposed work can illustrate the high accuracy of its models by offering a comparison table or graph displaying the accuracy rates of the suggested models alongside other existing models. It can include a part describing the numerous tests and validations carried out to make sure the models function properly under various circumstances and with diverse datasets in order to highlight the models’ robustness. If the study provides a thorough explanation of how previously trained models were adjusted for the particular task at hand, it can highlight the application of transfer learning. Figures showing the transfer learning process can go along with this. The advantages of transfer learning, such as shortened training times and enhanced output. Moreover, the proposed system can be hosted on the Cloud and may become a useful technological assistant for health experts when screening tumors at an early stage. The proposed system’s limitations include its tendency to base assessments and predictions on training data and its potential inability to adapt its knowledge to novel circumstances. It is challenging to recognize and resolve problems like bias and lack of robustness. Increasing the transparency of AI systems using methods like explainable AI can aid in resolving these issues. Although there is room for improvement in the proposed system, there may not be any direction or assistance in putting the improvements into practice. It could be necessary to set up accountability and follow-up procedures to make sure the person acts in future work.

[1]  T. Hossain, F.S. Shishir, M. Ashraf, M.A. Al Nasim, F. Muhammad Shah, Brain tumor detection using convolutional neural network, 1st Int. Conf. Adv. Sci. Eng. Robot. Technol. 2019 (2019) 1–5, https://doi.org/10.1109/ ICASERT.2019.8934561. ICASERT 2019. [2]  R. Hua, et al., Segmenting brain tumor using cascaded V-nets in multimodal MR images, Front. Comput. Neurosci. 14 (2020) 1–20, https://doi.org/10.3389/ fncom.2020.00009. [3]  M. Agarwal, Optimized contrast enhancement for tumor detection, Int. J. Imaging Syst. Technol. Willey (2020) 1–17, https://doi.org/10.1002/ima.22408. [4]  A. Philips, D.L. Henshaw, G. Lamburn, M.J. O’Carroll, Authors’ comment on ‘brain tumours: rise in glioblastoma multiforme incidence in England 1995-2015 suggests an adverse environmental or lifestyle factor, J. Environ. Public Health 2018 (2018), https://doi.org/10.1155/2018/2170208. [5]  Z. Ur, M.S. Zia, G. Reddy, M. Yaqub, F. Jinchao, Texture based localization of a brain tumor from MR-images by using a machine learning approach, Med. Hypotheses, Elsevier 141 (2020) 1–12, https://doi.org/10.1016/j. mehy.2020.109705. [6]  S. Hasan, M. Yousif, T.M.J. Al-talib, Brain tumor classification using probabilistic neural network, J. Fundam. Appl. Sci. 10 (2018) 667–670. [7]  S. Arivoli, K.J. Ravindran, R. Raveen, S. Tennyson, Detection and classification of brain tumor using machine learning approaches, Int. J. Res. Pharm. Sci. 10 (3) (2019) 2153–2162. [8]  N.V. Shree, Identification and classification of brain tumor MRI images with feature extraction using DWT and probabilistic neural network, Brain Informatics, Springer 5 (2018), https://doi.org/10.1007/s40708-017-0075-5. [9]  A. Rehman, S. Naz, M.I. Razzak, F. Akram, M. Imran, A deep learning-based framework for automatic brain tumors classification using transfer learning, Circuits, Syst. Signal Process. Birkhauser 39 (2) (2020) 757–775, https://doi.org/ 10.1007/s00034-019-01246-3. [10]  Z.N.K. Swati, et al., Brain tumor classification for MR images using transfer learning and fine-tuning, Comput. Med. Imaging Graph. Elsevier 75 (2019) 34–46, https://doi.org/10.1016/j.compmedimag.2019.05.001.