Quantum-Inspired Adaptive Feature Fusion for Highly Accurate Brain Tumor Classification in MRI Images

Abstract

Accurate brain tumor classification from magnetic resonance imaging (MRI) is crucial for early diagnosis and effective treatment planning. Many deep learning models rely on simple feature concatenation, which limits the use of complementary information from different feature types. This work proposes a Quantum-Inspired Adaptive Feature Fusion (QIAFF) method integrated with deep learning for improved brain tumor classification. MRI images are preprocessed using median and anisotropic diffusion filtering to enhance image quality. Handcrafted features such as HOG and texture descriptors, along with CNN deep features, are extracted to represent tumor characteristics. Quantum-inspired probability principles are used to assign adaptive weights to each feature set through a rotation-based update rule. The weighted features are fused into an optimal representation before classification. A deep neural network then performs tumor classification using the fused features. Experimental results show improved accuracy, precision, recall, and F1-score compared to conventional fusion and deep learning methods. The proposed approach enhances feature discrimination and generalization, making it effective for reliable brain tumor detection from MRI images.