Hybrid Deep Learning Algorithm for Multi-Grade Brain Tumor Classification

One of the potentially fatal conditions impacting a person's general health and brain function is a brain tumor. Early-stage brain tumor identification and classification accuracy is critical to saving lives. One of the most popular methods used in the medical field to classify brain tumors is deep learning. Nevertheless, current deep learning methods are insufficiently effective in precisely categorizing the various stages of brain cancers. In this study, Ensemble Learning with Deep Convolutional Long Short-Term Memory (EL-DCLSTM) was proposed to categorize multiple grades of brain tumors accurately. The first step in the suggested methodology is the acquisition and preparation of brain MRI data. As a result, the enhanced U-Net method was used for the segmentation stage in order to extract the region of interest from the previously processed images. In addition, the Firefly-optimized ResNet architecture was used for feature extraction, which involved selecting and extracting the most pertinent features for classifier training from the segmented images. The suggested EL-DCLSTM was applied for brain tumor classification after feature extraction. The system can accurately manage fluctuations in MRI data thanks to the DCLSTM design, which combines the effectiveness of convolutional and LSTM layers for capturing both spatial and temporal properties in the MRI images. However, ensemble learning generates the final classification results by aggregating the predictions from individual DCLSTM models that were trained on the extracted feature set's feature subsets. The suggested approach obtained better accuracy of 0.98172 and 0.99138 for 70% and 80% training ratios, according to the experimental results.