Optimized Quantum CNN With Improvised iHow Algorithm for MRI-Based Dementia Diagnosis

Dementia, a neurodegenerative disorder, requires early prediction and effective diagnosis for providing better treatment to avert the loss. The detection and classification of disease is intricate and needs optimized solutions using deep learning techniques. In context with this, the proposed work presents a novel approach with the inclusion of an Optimized Quantum Convolutional Neural Network (Q-CNN). The optimization algorithm utilized is the Improved iHow Optimization Algorithm (IiHOA), which fine-tunes the hyper-parameters in the proposed Q-CNN. Moreover, the MRI images are pre-processed with the combined act of B-spline and Lanczos Interpolation by removing the noises and preserving the sensitive details. This enhances the image quality and can be used to extract features. For feature extraction, U-Net is used, and the decomposition of MRI images into multi-scale frequency components is done for extracting the spatial and spectral information related to brain abnormalities. For further refining the extracted features, the Adaptive Hiking Optimization algorithm (AHOA) is used, which selects the appropriate features and therein significantly mitigates the computational complexities and enhances the classification accuracy. The feature representation and non-linear patterns from the MRI images are utilized by the proposed QCNN for the classification. The benchmark OASIS dementia dataset attained an accuracy of 98%, and various parameters were analysed with the proposed QCNN. This ensures a highly accurate dementia classification with the proposed work.