Journal of Theoretical and Applied Information Technology
QUANTUM-ENHANCED EYE-TRACKING BIOMARKERS TO IMPROVE EARLY DETECTION AND DIGITAL PHENOTYPING OF AUTISM SPECTRUM DISORDER: A HYBRID QUANTUM–CLASSICAL LEARNING APPROACH
SUNEETHA DAVULURI, ANJI REDDY TANGIRALA, VENUGOPAL BOPPANA, V P UMAMAHESWARA RAO MALLA, SREENIVAS VELAGAPUDI
Autism spectrum disorder (ASD) remains one of the foremost clinical challenges for early detection, and subjective diagnostic protocols constrain utilization while classical computational methods have failed to decode subtle oculomotor biomarker signatures. This work asks whether quantum feature encoding can move oculomotor ASD screening beyond the representational ceiling reached by classical eye-tracking classifiers, and answers this with a Hybrid Quantum–Classical Learning (QHCL) framework that combines parameterized quantum circuits with classical deep learning architectures to extract, encode, and classify eye-tracking biomarkers for early ASD detection and digital phenotyping. This is achieved by encoding raw gaze signals — fixation durations, saccadic velocities, scan path geometries, and pupillary response dynamics — into quantum Hilbert space representations through amplitude embedding to enable simultaneous processing of high-dimensional spatiotemporal oculomotor patterns. Quantum kernel-based feature learning reveals latent inter-feature correlations elusive to standard algorithms, and an entanglement-driven circuit layer forms enhanced patient-specific gaze embeddings. These embeddings are passed into a classical dual-head classifier trained for binary ASD classification and continuous digital phenotype evaluation along severity spectra. On three benchmark paediatric eye-tracking datasets, the proposed framework achieves classification accuracies of 95.8%, 97.4% and 96.2%, outperforming the strongest available classical baseline by 1.3–3.5 percentage points and improving quantum-versus-PCA feature learning by up to 9.3% at the individual feature-family level. The research contribution of this work is twofold: (i) it is the first study to apply quantum amplitude encoding and entanglement-driven parameterised circuits to oculomotor biomarker analysis for ASD, establishing that quantum Hilbert-space representations capture inter-feature correlations among fixation, saccadic, scanpath and pupillary signals that classical PCA-based pipelines discard; and (ii) it introduces a shared dual-head quantum-classical architecture that jointly optimises binary detection and four-class severity phenotyping from a single embedding, narrowing the accuracy gap that typically separates these two tasks in the literature. These findings open a new, non-invasive and computationally efficient pathway toward precision ASD screening with clinical applicability.
Autism spectrum disorder (ASD) remains one of the foremost clinical challenges for early detection, and subjective diagnostic protocols constrain utilization while classical computational methods have failed to decode subtle oculomotor biomarker signatures. This work asks whether quantum feature encoding can move oculomotor ASD screening beyond the representational ceiling reached by classical eye-tracking classifiers, and answers this with a Hybrid Quantum–Classical Learning (QHCL) framework that combines parameterized quantum circuits with classical deep learning architectures to extract, encode, and classify eye-tracking biomarkers for early ASD detection and digital phenotyping. This is achieved by encoding raw gaze signals — fixation durations, saccadic velocities, scan path geometries, and pupillary response dynamics — into quantum Hilbert space representations through amplitude embedding to enable simultaneous processing of high-dimensional spatiotemporal oculomotor patterns. Quantum kernel-based feature learning reveals latent inter-feature correlations elusive to standard algorithms, and an entanglement-driven circuit layer forms enhanced patient-specific gaze embeddings. These embeddings are passed into a classical dual-head classifier trained for binary ASD classification and continuous digital phenotype evaluation along severity spectra. On three benchmark paediatric eye-tracking datasets, the proposed framework achieves classification accuracies of 95.8%, 97.4% and 96.2%, outperforming the strongest available classical baseline by 1.3–3.5 percentage points and improving quantum-versus-PCA feature learning by up to 9.3% at the individual feature-family level. The research contribution of this work is twofold: (i) it is the first study to apply quantum amplitude encoding and entanglement-driven parameterised circuits to oculomotor biomarker analysis for ASD, establishing that quantum Hilbert-space representations capture inter-feature correlations among fixation, saccadic, scanpath and pupillary signals that classical PCA-based pipelines discard; and (ii) it introduces a shared dual-head quantum-classical architecture that jointly optimises binary detection and four-class severity phenotyping from a single embedding, narrowing the accuracy gap that typically separates these two tasks in the literature. These findings open a new, non-invasive and computationally efficient pathway toward precision ASD screening with clinical applicability.
Autism Spectrum Disorder (ASD) is a neurodevelopmental condition whose diagnosis relies heavily on subjective behavioral assessments. Electroencephalography (EEG) provides an inexpensive and non-invasive method of objective biomarker detection but is plagued by high dimensionality and noise that complicates automated analysis. This article introduces a hybrid Transformer–Quantum Neural Network (QNN) model that combines deep learning and quantum computing to enhance P300 targeted ASD responses from non-targeted responses. The model uses transformer capabilities captures cross-domain correlations and feature-level dependencies across heterogeneous feature sources and the quantum heads introduced via parameterized quantum circuits detects high-order, nonlinear relations in a Hilbert space via entanglement and superposition. Experimental results on the BCIAUT-P300 dataset better classical alternatives with higher diagnostic accuracy at higher accuracy (0.921) and F1-score (0.915). Statistical comparisons confirm that the hybrid approach yields significant improvements in discriminative reliability and diagnostic odds ratio, justifying quantum feature embedding’s expressivity and efficiency. The results bear witness to the feasibility and potential of hybrid quantum–classical systems as potential future intelligent diagnostic devices for ASD responses from EEG signals.
Early and accurate identification of autism spectrum disorder (ASD) remains a central challenge in clinical neuroscience due to the subjective nature and high resource demands of traditional diagnostic methods. This paper presents a hybrid classical–quantum learning framework for ASD detection, combining the representational power of quantum circuits with the reliability of classical machine learning. The proposed model, a Variational Quantum Classifier (VQC) implemented using PennyLane and PyTorch, processes structured behavioural and demographic screening data from child, adolescent, and adult cohorts. Features are classically preprocessed, encoded into quantum states through angle embedding, and refined using strongly entangling layers, before being post-processed through a classical dense network. Experimental results show perfect classification performance with 100% accuracy, precision, recall, and F1-score, outperforming comparable classical approaches. The training phase exhibited smooth convergence across 70 epochs, confirming model stability and learning efficiency. Although evaluated in simulation, this study establishes a proof-of-concept for leveraging quantum machine learning in medical diagnostics, laying groundwork for future integration with real quantum hardware in clinical screening applications.
Autism Spectrum Disorder (ASD) is a heterogeneous neurodevelopmental disorder; classical machine learning approaches, using behavioral, neuroimaging, and electrophysiological data, find it difficult to detect the complex interdependencies between the nonlinear characteristics of and temporal change in risks. This research paper introduces the Adaptive Quantum-Classical Feature Learning (AQ-CFL) approach, which incorporates a quantum feature map that captures non-linear correlations and a classical deep neural network for cross-sectional ASD risk classification and longitudinal prediction of future behavioral patterns. Classical features are mapped into the quantum Hilbert space using entangling angle encoding, processed using a variational quantum classifier, and merged with classical embeddings using the attention-gated layer; a transformer block learns the longitudinal visits to predict future severity and adaptive behaviors. AQ-CFL is formulated mathematically in terms of the feature map, parameter-shift gradient computation rule, loss, and computational complexity compared to classical solutions. The feasibility of the proposed approach is confirmed by state-vector simulation using the features obtained from the ABIDE database and demonstrating slight but consistent improvement over classical methods and similar trajectory prediction.
The classification of Autism Spectrum Disorder (ASD) using neuroimaging techniques is still not an easy task, as the features have a high dimension, inter-site variability, and a small number of labeled samples. In this paper, HybridVQC, a quantum-inspired hybrid neural architecture, incorporates mathematical concepts of quantum circuits into a fully classical, learning architecture that can be executed on a GPU. ABIDE Structural MRI slices are fed through a pretrained backbone of ResNet-18 to obtain deep representations, and the deep representations are further shrunk to 16 principal components in Principal Component Analysis (PCA). The low features are subsequently applied to a special QuantumLikeLayer, which uses trigonometric encoding and dense mixing of features in order to simulate quantum rotation and entanglement effects on ordinary CUDA. The results of experiments with 1,693 structural MRI slices reveal that the maximum validation and test accuracy is 80.63 and 75.0, respectively, versus 56% of a classical SVM baseline using the same parameters. The findings point to quantum-inspired non-linear transformations have the potential to enhance feature separability and training stability in neuroimaging classification, and do not use quantum simulators or physical quantum systems.
The classification of Autism Spectrum Disorder (ASD) using neuroimaging techniques is still not an easy task, as the features have a high dimension, inter-site variability, and a small number of labeled samples. In this paper, HybridVQC, a quantum-inspired hybrid neural architecture, incorporates mathematical concepts of quantum circuits into a fully classical, learning architecture that can be executed on a GPU. ABIDE Structural MRI slices are fed through a pretrained backbone of ResNet-18 to obtain deep representations, and the deep representations are further shrunk to 16 principal components in Principal Component Analysis (PCA). The low features are subsequently applied to a special QuantumLikeLayer, which uses trigonometric encoding and dense mixing of features in order to simulate quantum rotation and entanglement effects on ordinary CUDA. The results of experiments with 1,693 structural MRI slices reveal that the maximum validation and test accuracy is 80.63 and 75.0, respectively, versus 56% of a classical SVM baseline using the same parameters. The findings point to quantum-inspired non-linear transformations have the potential to enhance feature separability and training stability in neuroimaging classification, and do not use quantum simulators or physical quantum systems.
Early detection of Autism Spectrum Disorder (ASD) is crucial for initiating interventions that can significantly improve developmental outcomes. Traditional diagnostic methods are often subjective and time-consuming, leading to delays in diagnosis. While classical machine learning models have been applied to ASD detection with promising results, they face challenges in processing complex, high-dimensional medical data. Quantum computing utilizes superposition, and entanglement offers a novel approach to handle such data more efficiently. This study investigates the application of Quantum Support Vector Machines (QSVM) for ASD detection using the Autism Screening Data for Toddlers (ASDTests) dataset. After preprocessing and feature selection, we trained a QSVM model. We compared its performance with classical machine learning models, including Support Vector Machine (SVM), Random Forest (RF), K-Nearest Neighbors (KNN), and Naive Bayes (NB). We evaluated the models based on accuracy, precision, recall, and F1-score across multiple random subsets to assess robustness. The QSVM model achieved the highest mean accuracy (96.7 %), precision (97.2 %), and F1-score (95.9 %) with low variance, indicating strong and consistent performance. Statistical analyses using the Paired t-test and Friedman Test showed marginal improvements over the SVM model (p = 0.0495), with no significant difference compared to RF and NB models. These results suggest that QSVM can enhance the accuracy and robustness of ASD detection compared to classical models. Future work could explore hybrid models combining quantum and classical machine learning approaches to improve performance. Applying these methods to different ASD datasets may provide broader validation and insights into their generalizability.