Enhanced Brain Tumor Detection using Adaptive Preprocessing and Hybrid Deep Learning Techniques

Abstract

Neurological and brain cancers represent major global health issues, with MRI serving as a vital diagnostic tool. However, interpreting MRI images manually is often slow and prone to inconsistency. This study presents a sophisticated framework for detecting brain tumors using Dynamic Contrast-Enhanced MRI (DCE-MRI). The framework integrates advanced preprocessing and hybrid deep learning techniques to enhance performance. Adaptive multi-scale Gaussian filtering is used initially to reduce noise and enhance picture quality. Following the identification of the most relevant features using Recursive Feature Elimination with Cross-Validation (RFECV), Levy Flight Particle Swarm Optimization (LFPSO) is used to optimize feature extraction. Convolutional neural networks (CNNs) and long short-term memory networks (LSTMs) are combined in a hybrid deep learning model to accomplish the classification, which successfully captures both temporal and spatial patterns. This methodology offers notable improvements in detection accuracy and processing speed, showcasing its potential for clinical use and better patient outcomes.