Brain Tumor Detection and Classification Using a New Evolutionary Convolutional Neural Network

Amin Abdollahi Dehkordi,Mina Hashemi,Mehdi Neshat,Seyedali Mirjalili,Ali Safaa Sadiq
DOI: https://doi.org/10.48550/arXiv.2204.12297
2022-04-26
Abstract:A definitive diagnosis of a brain tumour is essential for enhancing treatment success and patient survival. However, it is difficult to manually evaluate multiple magnetic resonance imaging (MRI) images generated in a clinic. Therefore, more precise computer-based tumour detection methods are required. In recent years, many efforts have investigated classical machine learning methods to automate this process. Deep learning techniques have recently sparked interest as a means of diagnosing brain tumours more accurately and robustly. The goal of this study, therefore, is to employ brain MRI images to distinguish between healthy and unhealthy patients (including tumour tissues). As a result, an enhanced convolutional neural network is developed in this paper for accurate brain image classification. The enhanced convolutional neural network structure is composed of components for feature extraction and optimal classification. Nonlinear Lévy Chaotic Moth Flame Optimizer (NLCMFO) optimizes hyperparameters for training convolutional neural network layers. Using the BRATS 2015 data set and brain image datasets from Harvard Medical School, the proposed model is assessed and compared with various optimization techniques. The optimized CNN model outperforms other models from the literature by providing 97.4% accuracy, 96.0% sensitivity, 98.6% specificity, 98.4% precision, and 96.6% F1-score, (the mean of the weighted harmonic value of CNN precision and recall).
Neural and Evolutionary Computing
What problem does this paper attempt to address?
The problem that this paper attempts to solve is to improve the accuracy of brain tumor detection and classification through an improved Convolutional Neural Network (CNN). Specifically, the goal of the paper is to use brain MRI images to distinguish between healthy and diseased patients (including tumor tissues). To achieve this goal, the researchers developed an enhanced convolutional neural network structure, which includes feature extraction and optimal classification components, and uses the Non - Linear Levy Chaotic Moth - Flame Optimizer (NLCMFO) to optimize hyper - parameters for training the convolutional neural network layers. By using the BRATS 2015 dataset and the brain image dataset of Harvard Medical School to evaluate the proposed model and comparing it with various optimization techniques, the results show that the optimized CNN model is superior to other models in the literature in terms of accuracy, sensitivity, specificity, precision, and F1 - score. ### Main problems solved in the paper: 1. **Improve the accuracy of brain tumor detection**: Traditional methods of manually evaluating brain MRI images are time - consuming and error - prone, so a more accurate computer - aided diagnosis method is required to automatically detect and classify brain tumors. 2. **Optimize the hyper - parameters of CNN**: By introducing the NLCMFO algorithm to optimize the hyper - parameters of CNN, the training efficiency and performance of the model are improved. 3. **Handle large - scale datasets**: Use large - scale MRI datasets (such as BRATS 2015) to verify the effectiveness of the model and compare it with other optimization methods. ### Specific methods and techniques: - **Enhanced Convolutional Neural Network (CNN)**: A new CNN structure is designed, including feature extraction and classification components. - **Non - Linear Levy Chaotic Moth - Flame Optimizer (NLCMFO)**: Used to optimize the hyper - parameters of CNN and improve the training effect of the model. - **Datasets**: The BRATS 2015 dataset and the brain image dataset of Harvard Medical School are used for model evaluation. ### Experimental results: - **Accuracy**: 97.4% - **Sensitivity**: 96.0% - **Specificity**: 98.6% - **Precision**: 98.4% - **F1 - score**: 96.6% These results indicate that the proposed model performs well in brain tumor detection and classification tasks and has high accuracy and robustness.