Hybrid Deep Learning Model for Enhanced Brain Tumor Classification Using VGG19, LSTM, and SVM Algorithms

Brain tumors are an important worldwide health issue, contributing to high rates of morbidity and mortality. Their early detection is vital for enhancing patient outcomes; yet, it remains a tough undertaking despite ongoing improvements in medical technology. This research offers a unique hybrid Deep Learning model aiming at enhancing the efficiency of brain tumor categorization. The model incorporates the characteristics of VGG19 for the feature extraction process, LSTM for understanding temporal correlations, and SVM for accurate classification. The dataset utilized for training and evaluation consists of 7023 MRI images sourced from three independent datasets: Figshare, SARTAJ, and Br35H. An important enhancement includes rectifying misclassifications in the glioma category by replacing mislabeled SARTAJ photos with appropriately classified images from Figshare. The suggested model's performance is tested using numerous measures, including Precision, Recall, F1-Score, and Accuracy. The results obtained by the model have a classification accuracy of 97%, making it highly useful for brain tumor detection and diagnosis.