A Quantum-Inspired Attention Network with Orthogonal Polynomials for Medical Image Classification

In this research, an Orthogonal Polynomials Transformation and Quantum Inspired Attention Network (QIAN) based medical image classification framework called QuAntNet is presented. It integrates low level image features are automatically extracted from brain MRI images with expert level clinical insights to identify diagnosis precise classes. The proposed method extracts fusion edge, texture and deep level features from the image based on OPT and Convolutional Neural Network techniques. The extracted classical features are submitted to quantum-inspired attention network mechanism for analyzing intricate relationships between different discriminative image features. The QuAntNet is validated on reputed brain MRI datasets namely BraTS2020 and Figshare, which includes different classes of brain tumors. Experiment conducted on BraTS2020 dataset, the QuAntNet method establishes overall accuracy upto 98.7%, sensitivity upto 98.2%, specificity upto 99.0% and F1-score of 98.4%. Also, when valiadated on Figshare, it produces accuracy of 99.3%, sensitivity of 98.9%, specificity of 99.6% and F1-score of 99.1%. Experimental outcomes clearly validates that there is significant enhancement in classification accuracy incorporating fusion edge, texture and deep features and QIAN methods enhances the classification perfomance in the QuAntNet framework.