Novel Adaptive Preprocessing and Hybrid Deep Learning Model for Early Brain Tumor Detection

Abstract

Early detection of brain tumors is critical for improving patient survival rates and treatment planning. Traditional diagnostic methods are time-consuming and prone to human error. This paper proposes a novel adaptive pre-processing framework combined with a hybrid CNN–ResNet model for accurate and efficient brain tumor detection using MRI images. The proposed system integrates adaptive noise removal, contrast enhancement, and normalization techniques to improve image quality before feeding into a hybrid deep learning model. The architecture combines the feature extraction capability of Convolutional Neural Networks (CNN) with the deep residual learning of ResNet, enabling improved detection accuracy and reduced vanishing gradient issues. Experimental results demonstrate that the novel adaptive pre-processing and proposed hybrid CNN–ResNet model, improve early detection accuracy by integration of preprocessing techniques and residual learning enhances feature extraction, improves generalization, and reduces vanishing gradient problems. Hybrid model achieves superior performance compared to traditional CNN and standalone ResNet models, making it highly suitable for early-stage tumor detection.