Hybrid Quantum-Classical Neural Network for Diagnosis of Autism Spectrum Disorder

Autism Spectrum Disorder (ASD) affects almost 3% of children worldwide, and due to the complexity and heterogeneity of ASD, accurate and efficient diagnosis is difficult to achieve. Qualitative diagnosis based on observation of behavior has a major risk of misdiagnosis. Meanwhile, the most efficient quantitative algorithms in modern literature are similarly problematic, with misdiagnosis rates of up to 44 %, execution times of over 7 hours, and all algorithms are purely classical. This project is an implementation of a hybrid quantum-classical neural network (HQCNN) to improve upon classical neural networks by applying quantum algorithms while only using fMRI data. The proposed neural network implements a variational quantum deflation (VQD) algorithm, an extension of the variational quantum eigensolver (VQE) framework for finding the first <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$k$</tex> excited states of a Hamiltonian. This study shows that the VQD algorithm produces eigenvalues effectively equivalent to the classically computed ones, and the general algorithm has up to 70% accuracy with less than an hour of execution time, a major improvement from the 56% accuracy and over 7 hour runtimes of most classical algorithms. With the development of quantum computing technology, the VQD algorithm will continue to be optimized. Furthermore, the HQCNN is generally scalable, and only requires fMRI data for analysis. This study shows a faster and more accurate diagnosis of ASD using quantum algorithms.