Challenges
Datasets
Workspaces
Papers
Discussions
Leaderboard
Log inSign up
Challenges
Datasets
Workspaces
Papers
Discussions
Leaderboard
Blog
Job Board
Q3AS

© 2026 Aqora Quantum S.A.S.

TermsPrivacyLegal Notice

International Conference on Electronics, Computing, Communication and Control Technology (ICECCC)

HQFL-Net: A Hybrid Quantum-Classical Federated Learning Framework for Explainable Cardiovascular Disease Prediction on Edge Devices

Sahana Edwin, Pallabee Padhi, Thalaimalaichamy M, Surendar Madhawa

Published Jul 1, 2026• 0 citations
Publisher

Abstract

Cardiovascular disease accounts for roughly 32% of all global deaths; however, popular risk scores like Framingham have a sensitivity of around 68-72 %, making a need for accurate prediction that is privacy-preserving and deployable at the point-of-care necessary. This task is made difficult by regulation (HIPAA/GDPR), non-IID datasets, and interpretability concerns in decentralized healthcare scenarios. Federated Learning (FL) algorithms can protect privacy but are poor at capturing non-linear dependencies between features, while Quantum Machine Learning (QML) can represent such complex relationships but is not applied in federated nor edge settings. This work introduces HQFL-Net, a hybrid quantumclassical federated learning architecture whereby edge devices train a classical XGBoost model together with a quantum model represented by a shallow 4-qubit variational quantum circuit, employing entropy-weighted aggregation to improve convergence on non-IID data, as well as classical surrogacy (R2=0.97) to generalize Shapley Additive Explanations to quantum features. On the Kaggle CVD dataset (70,000 examples) with edge emulation using Raspberry Pi 4 microcontrollers, HQFL-Net achieves F 1=0.92 ± 0.01 and A U C=0.94 ± 0.01, outperforming classical FedAvg (F 1=0.87) and centralized XGBoost (F 1 = 0.89) while being deployed with a 40 ms per sample latency on 5 W devices.

DOI
10.1109/iceccc70334.2026.11633013

Related papers

Explainable Quantum-Enhanced Federated Learning Framework for Real-Time Cardiovascular Risk Assessment

Sriramkumar R, Nagarjuna Naik Mude, R Rakshita +3 more

According to the World Health Organization, cardiovascular diseases are responsible for several million deaths worldwide each year, thus highlighting a prime necessity for developing methods of diagnosis that are accurate and rapid. Although recent advancements in quantum-enhanced machine learning have demonstrated enhanced prediction performance, current research is still limited by the reliance on static datasets, the lack of interpretability, and the lack of privacy-preserving mechanisms for data sharing across institutions. In this paper, we introduce an Explainable Quantum-Enhanced Federated Learning Framework for Real-Time Cardiovascular Risk Assessment that integrates the scalability of federated learning with the computational benefits of quantum models to allow multiple healthcare centers and remote devices to train models collaboratively without sharing raw patient data. Quantum-inspired classifiers such as variational circuits and Quantum SVM are integrated into the federated pipeline to accelerate convergence and improve robustness. An explanation layer was implemented to enhance clinical trust by applying changed feature attribution methods that help the physicians comprehend how much the important factors like cholesterol, blood pressure, and ECG patterns contribute to the risk predictions. In addition, the system is meant for constant watch by the mixing of electronic health records and wearable sensor streams, thus making it possible to switch from one-time prediction to dynamically and personally tailored risk assessment. The proposed technique has been validated through experiments on standard heart disease datasets where it is also said to have done the opposite: it has improved predictive accuracy, reduced communication overhead compared to traditional federated systems, and provided insights that are ready for supporting clinical decision-making in real practice. Consequently, this framework is a first step towards quantum-driven healthcare solutions that are safe, scalable, and explainable.

International Conference on Next-Gen Quantum and Advanced Computing: Algorithms, Security, and Beyond (NQComp)•Mar 5, 2026•1 citation

Hybridized Quantum Computing and Federated Learning for Multi-Disease Prediction

M. S. Anbarasi, Aaron Pushparaj M. M., Boomika S. +1 more

The growing occurrence of patients with multiple chronic diseases has increased the need for efficient multidisease risk prediction systems in healthcare. However, clinical data is often distributed in various cloud, but it lacks in efficient privacy preservation. To address these challenges, this project proposes a hybridized framework for privacy-aware multi-disease risk prediction in distributed healthcare environments. nA key challenge in multi-disease data collection from heterogeneous clinical data may lag in normalization. So, cleaning and preprocessing is done on the first phase. Second challenge lies in capturing complex and non-linear relationships among clinical features within reduced feature spaces. To handle this, quantum-inspired feature encoding and variational quantum circuits are employed to enhance feature representations. Since data is distributed across cloud data privacy and confidentiality issues exist. To overcome this issue Federated Learning for multi-disease prediction is applied in the third phase. To address these challenges, this paper proposes a novel framework titled Hybridized Quantum Computing and Federated Learning for Multi-Disease Prediction (HQCF-MDRP). The proposed system integrates structured data preprocessing, quantum-assisted feature representation, and federated collaborative learning into a unified architecture. Initially, healthcare datasets are cleaned, normalized, and subjected to feature selection to improve data quality and reduce dimensionality. Selected features are then transformed using quantum-inspired angle encoding and Variational Quantum Circuits (VQC), enabling enhanced representation of hidden correlations among clinical attributes. Subsequently, Federated Learning with FedProx optimization is employed to train predictive models collaboratively across distributed healthcare institutions without sharing raw patient data.

International Journal of Innovative Science and Research Technology (IJISRT)•May 9, 2026•0 citations

DAQFL: Dynamic Aggregation Quantum Federated Learning Algorithm for Intelligent Diagnosis in Internet of Medical Things

Zhiguo Qu, Xuemeng Zhao, Le Sun +1 more

Federated learning (FL) is a privacy-preserving alternative to centralized machine learning, where model training is performed on local devices and only global model updates are shared, effectively addressing challenges, such as data silos and privacy protection. Recently, quantum FL (QFL), an emerging FL branch, has garnered significant attention in many industry applications. However, existing QFL algorithms predominantly employ average weighting for global model training, which shows poor performance on heterogeneous healthcare data. To address this challenge, this study proposes a dynamic aggregation QFL algorithm (DAQFL) for intelligent diagnosis. Specifically, it utilizes quantum neural networks (QNNs) as local training models and designs corresponding variational quantum circuits (VQC). To mitigate performance degradation caused by the heterogeneity of medical industrial data, a dynamic aggregation method based on accuracy is proposed to enhance global model performance effectively. Extensive experiments with three distribution settings, including independent and identically distributed (IID), non-independent and identically distributed (Non-IID), and long-tail datasets, show that DAQFL outperforms baseline algorithms in accuracy and training speed. It also performs well in privacy protection and robustness of anti-noise, improving its suitability for real-world medical applications.

IEEE Internet of Things Journal•Jan 31, 2025•28 citations

A Hybrid Quantum-Classical Machine Learning Framework for Early and Accurate Diagnosis of Chronic Diseases

Al Khan

Chronic diseases—such as diabetes, cardiovascular disorders, and chronic respiratory conditions—account for over 70% of global deaths annually, with late diagnosis being a primary contributor to poor outcomes. While machine learning (ML) models have shown promise in early detection, they often suffer from limited generalizability, data heterogeneity, and insufficient interpretability in clinical settings. This paper introduces a novel hybrid quantum-classical machine learning (HQML) framework that synergistically combines the pattern recognition power of classical deep neural networks with the high-dimensional optimization capabilities of quantum algorithms to enhance diagnostic accuracy, robustness, and early signal detection in chronic disease prediction. Using multimodal electronic health record (EHR) data—including clinical, genomic, and lifestyle variables—we train a quantum-enhanced feature selector followed by a classical interpretable classifier (e.g., SHAP-augmented XGBoost). The quantum component leverages a variational quantum circuit to identify non-linear, high-order feature interactions that classical methods often miss. Validated on datasets from the National Health and Nutrition Examination Survey (NHANES) and UK Biobank, our model achieves 94.7% AUC for type 2 diabetes prediction five years before clinical diagnosis—outperforming state-of-the-art baselines by 6.2%. Crucially, the framework maintains interpretability through post-hoc explainability and reduces data bias via fairness-aware quantum embedding. This research bridges quantum computing and public health, offering a scalable, ethically grounded diagnostic paradigm. By enabling earlier, more accurate, and equitable predictions, the HQML framework has significant potential to transform preventive care and reduce the global burden of chronic disease.

•Jan 7, 2026•0 citations

AF-HQCNN: Adaptive Federated Hybrid Quantum Convolutional Neural Network for Privacy-Preserving Medical Image Classification

Velkumar K, Dr.Mathalairaj J, Bhavani M +3 more

Medical image classification is an important task in clinical decision support systems. Although classical deep learning models can represent massive amounts of data, they are not very scalable DEEP LEARNING REVIEW and stand each in their own right for privacy loss and computation inefficiency. Quantum machine learning (QML) has been proposed as an exciting field to complement classical machine learning by using quantum superposition and entanglement for increasing the expressibility of the models. Since the hybrid quantum-classical methods are subject to high qubit complexity [10], training instability (barren plateau problem) [13, 14], non-existent privacy protection mechanisms and poor generalisability over trained datasets [7]. Herein, we propose AF-HQCNN: an Adaptive Federated Hybrid Quantum Convolutional Neural Network (1) a lightweight classical feature extractor using only MobileNetV2, (2) variational quantum circuit (VQC) with angle and amplitude encoding with 4-8 qubits, (3) novel adaptive momentum-based quantum optimizer to mitigate gradient vanishing and gradient explosion issues, (4) federated learning aggregation layer layer to ensure protection of patient data privacy,(5), multi-dataset validation on MedMNIST, HAM10000, Brain MRI & Chest X-Ray datasets. Our experimental results show that AF-HQCNN leads to the mean classification accuracy of 97.6%, outperforming existing hybrid models by a margin of 2.7% with 87% fewer trainable parameters, faster convergence and robustness against privacy attacks.

International Research Journal on Advanced Science Hub•Jul 24, 2026•0 citations

Quantum-Inspired Feature Selection and Federated Intelligence for Predictive Geriatric Healthcare

Mustafa Nawaz SM, Nithya Devi S M, Atyam Naga VSD Sri Veena +3 more

As the elderly population continues to grow, there is an urgent need for decentralized, privacy-preserving, and high-accuracy predictive healthcare models that can proactively manage chronic illnesses. Most traditional Federated Learning (FL) approaches often have challenges with non-IID data distribution, slow convergence rates, the potential for security attacks, and high communication cost. To capitalize on the benefits of FL while addressing limitations, we propose a Quantum-Inspired Federated Learning (QIFL) framework that combines quantum computing concepts with FL to improve predictive performance, security, and computational efficiency. The QIFL model utilizes Quantum Variational Circuits (QVCs) and Quantum-Inspired Stochastic Gradient Descent (QSGD) to expedite local model updates, while Quantum Differential Privacy (Q-DP) guarantees patient data protection. In addition, the framework employs a Quantum Gate-Based Feature Selection (QGFS) mechanism to identify informative biomarkers for early detection of age-related diseases such as cardiovascular Diseases, Alzheimer's and Diabetes. Additionally, a Quantum-Inspired Adaptive Federated Averaging (QAFedAvg) method is proposed to optimize the aggregation of the global model from heterogeneous healthcare nodes. The framework also uses Quantum Secure Homomorphic Encryption (Q-SHE) and Quantum-Key Distribution (QKD) as security and privacy approaches to prevent adversarial model poisoning attacks during a federated update. The QIFL framework is evaluated on real-world elderly healthcare datasets (MIMIC-III, eICU, and PhysioNet) and performance is based on prediction accuracy, convergence time, adversarial robustness, and privacy leakage resistance. The QIFL framework demonstrates superior performance over standard FL frameworks, achieving a prediction accuracy of$97.8 \% (\pm 0.3 \%)$, a convergence time of$\mathbf{5 2}$percent quicker, a$\mathbf{4 1 \%}$lower adversarial success rate, and a reduced communication cost by 42%. This work reveals the effectiveness of quantum inspired computing and federated learning methods to produce an accurate, secure, and computationally efficient predictive healthcare model for the elderly population. The authors will continue to explore future work using Quantum Neural Networks (QNN) to enhance predictive intelligence and real-time diagnostics.

IEEE International Conference on Computing, Power and Communication Technologies (IC2PCT)•May 15, 2026•0 citations
Cardiovascular disease accounts for roughly 32% of all global deaths; however, popular risk scores like Framingham have a sensitivity of around 68-72 %, making a need for accurate prediction that is privacy-preserving and deployable at the point-of-care necessary. This task is made difficult by regulation (HIPAA/GDPR), non-IID datasets, and interpretability concerns in decentralized healthcare scenarios. Federated Learning (FL) algorithms can protect privacy but are poor at capturing non-linear dependencies between features, while Quantum Machine Learning (QML) can represent such complex relationships but is not applied in federated nor edge settings. This work introduces HQFL-Net, a hybrid quantumclassical federated learning architecture whereby edge devices train a classical XGBoost model together with a quantum model represented by a shallow 4-qubit variational quantum circuit, employing entropy-weighted aggregation to improve convergence on non-IID data, as well as classical surrogacy (R2=0.97) to generalize Shapley Additive Explanations to quantum features. On the Kaggle CVD dataset (70,000 examples) with edge emulation using Raspberry Pi 4 microcontrollers, HQFL-Net achieves F 1=0.92 ± 0.01 and A U C=0.94 ± 0.01, outperforming classical FedAvg (F 1=0.87) and centralized XGBoost (F 1 = 0.89) while being deployed with a 40 ms per sample latency on 5 W devices.
According to the World Health Organization, cardiovascular diseases are responsible for several million deaths worldwide each year, thus highlighting a prime necessity for developing methods of diagnosis that are accurate and rapid. Although recent advancements in quantum-enhanced machine learning have demonstrated enhanced prediction performance, current research is still limited by the reliance on static datasets, the lack of interpretability, and the lack of privacy-preserving mechanisms for data sharing across institutions. In this paper, we introduce an Explainable Quantum-Enhanced Federated Learning Framework for Real-Time Cardiovascular Risk Assessment that integrates the scalability of federated learning with the computational benefits of quantum models to allow multiple healthcare centers and remote devices to train models collaboratively without sharing raw patient data. Quantum-inspired classifiers such as variational circuits and Quantum SVM are integrated into the federated pipeline to accelerate convergence and improve robustness. An explanation layer was implemented to enhance clinical trust by applying changed feature attribution methods that help the physicians comprehend how much the important factors like cholesterol, blood pressure, and ECG patterns contribute to the risk predictions. In addition, the system is meant for constant watch by the mixing of electronic health records and wearable sensor streams, thus making it possible to switch from one-time prediction to dynamically and personally tailored risk assessment. The proposed technique has been validated through experiments on standard heart disease datasets where it is also said to have done the opposite: it has improved predictive accuracy, reduced communication overhead compared to traditional federated systems, and provided insights that are ready for supporting clinical decision-making in real practice. Consequently, this framework is a first step towards quantum-driven healthcare solutions that are safe, scalable, and explainable.
The growing occurrence of patients with multiple chronic diseases has increased the need for efficient multidisease risk prediction systems in healthcare. However, clinical data is often distributed in various cloud, but it lacks in efficient privacy preservation. To address these challenges, this project proposes a hybridized framework for privacy-aware multi-disease risk prediction in distributed healthcare environments. nA key challenge in multi-disease data collection from heterogeneous clinical data may lag in normalization. So, cleaning and preprocessing is done on the first phase. Second challenge lies in capturing complex and non-linear relationships among clinical features within reduced feature spaces. To handle this, quantum-inspired feature encoding and variational quantum circuits are employed to enhance feature representations. Since data is distributed across cloud data privacy and confidentiality issues exist. To overcome this issue Federated Learning for multi-disease prediction is applied in the third phase. To address these challenges, this paper proposes a novel framework titled Hybridized Quantum Computing and Federated Learning for Multi-Disease Prediction (HQCF-MDRP). The proposed system integrates structured data preprocessing, quantum-assisted feature representation, and federated collaborative learning into a unified architecture. Initially, healthcare datasets are cleaned, normalized, and subjected to feature selection to improve data quality and reduce dimensionality. Selected features are then transformed using quantum-inspired angle encoding and Variational Quantum Circuits (VQC), enabling enhanced representation of hidden correlations among clinical attributes. Subsequently, Federated Learning with FedProx optimization is employed to train predictive models collaboratively across distributed healthcare institutions without sharing raw patient data.
Federated learning (FL) is a privacy-preserving alternative to centralized machine learning, where model training is performed on local devices and only global model updates are shared, effectively addressing challenges, such as data silos and privacy protection. Recently, quantum FL (QFL), an emerging FL branch, has garnered significant attention in many industry applications. However, existing QFL algorithms predominantly employ average weighting for global model training, which shows poor performance on heterogeneous healthcare data. To address this challenge, this study proposes a dynamic aggregation QFL algorithm (DAQFL) for intelligent diagnosis. Specifically, it utilizes quantum neural networks (QNNs) as local training models and designs corresponding variational quantum circuits (VQC). To mitigate performance degradation caused by the heterogeneity of medical industrial data, a dynamic aggregation method based on accuracy is proposed to enhance global model performance effectively. Extensive experiments with three distribution settings, including independent and identically distributed (IID), non-independent and identically distributed (Non-IID), and long-tail datasets, show that DAQFL outperforms baseline algorithms in accuracy and training speed. It also performs well in privacy protection and robustness of anti-noise, improving its suitability for real-world medical applications.
Chronic diseases—such as diabetes, cardiovascular disorders, and chronic respiratory conditions—account for over 70% of global deaths annually, with late diagnosis being a primary contributor to poor outcomes. While machine learning (ML) models have shown promise in early detection, they often suffer from limited generalizability, data heterogeneity, and insufficient interpretability in clinical settings. This paper introduces a novel hybrid quantum-classical machine learning (HQML) framework that synergistically combines the pattern recognition power of classical deep neural networks with the high-dimensional optimization capabilities of quantum algorithms to enhance diagnostic accuracy, robustness, and early signal detection in chronic disease prediction. Using multimodal electronic health record (EHR) data—including clinical, genomic, and lifestyle variables—we train a quantum-enhanced feature selector followed by a classical interpretable classifier (e.g., SHAP-augmented XGBoost). The quantum component leverages a variational quantum circuit to identify non-linear, high-order feature interactions that classical methods often miss. Validated on datasets from the National Health and Nutrition Examination Survey (NHANES) and UK Biobank, our model achieves 94.7% AUC for type 2 diabetes prediction five years before clinical diagnosis—outperforming state-of-the-art baselines by 6.2%. Crucially, the framework maintains interpretability through post-hoc explainability and reduces data bias via fairness-aware quantum embedding. This research bridges quantum computing and public health, offering a scalable, ethically grounded diagnostic paradigm. By enabling earlier, more accurate, and equitable predictions, the HQML framework has significant potential to transform preventive care and reduce the global burden of chronic disease.
Medical image classification is an important task in clinical decision support systems. Although classical deep learning models can represent massive amounts of data, they are not very scalable DEEP LEARNING REVIEW and stand each in their own right for privacy loss and computation inefficiency. Quantum machine learning (QML) has been proposed as an exciting field to complement classical machine learning by using quantum superposition and entanglement for increasing the expressibility of the models. Since the hybrid quantum-classical methods are subject to high qubit complexity [10], training instability (barren plateau problem) [13, 14], non-existent privacy protection mechanisms and poor generalisability over trained datasets [7]. Herein, we propose AF-HQCNN: an Adaptive Federated Hybrid Quantum Convolutional Neural Network (1) a lightweight classical feature extractor using only MobileNetV2, (2) variational quantum circuit (VQC) with angle and amplitude encoding with 4-8 qubits, (3) novel adaptive momentum-based quantum optimizer to mitigate gradient vanishing and gradient explosion issues, (4) federated learning aggregation layer layer to ensure protection of patient data privacy,(5), multi-dataset validation on MedMNIST, HAM10000, Brain MRI & Chest X-Ray datasets. Our experimental results show that AF-HQCNN leads to the mean classification accuracy of 97.6%, outperforming existing hybrid models by a margin of 2.7% with 87% fewer trainable parameters, faster convergence and robustness against privacy attacks.
As the elderly population continues to grow, there is an urgent need for decentralized, privacy-preserving, and high-accuracy predictive healthcare models that can proactively manage chronic illnesses. Most traditional Federated Learning (FL) approaches often have challenges with non-IID data distribution, slow convergence rates, the potential for security attacks, and high communication cost. To capitalize on the benefits of FL while addressing limitations, we propose a Quantum-Inspired Federated Learning (QIFL) framework that combines quantum computing concepts with FL to improve predictive performance, security, and computational efficiency. The QIFL model utilizes Quantum Variational Circuits (QVCs) and Quantum-Inspired Stochastic Gradient Descent (QSGD) to expedite local model updates, while Quantum Differential Privacy (Q-DP) guarantees patient data protection. In addition, the framework employs a Quantum Gate-Based Feature Selection (QGFS) mechanism to identify informative biomarkers for early detection of age-related diseases such as cardiovascular Diseases, Alzheimer's and Diabetes. Additionally, a Quantum-Inspired Adaptive Federated Averaging (QAFedAvg) method is proposed to optimize the aggregation of the global model from heterogeneous healthcare nodes. The framework also uses Quantum Secure Homomorphic Encryption (Q-SHE) and Quantum-Key Distribution (QKD) as security and privacy approaches to prevent adversarial model poisoning attacks during a federated update. The QIFL framework is evaluated on real-world elderly healthcare datasets (MIMIC-III, eICU, and PhysioNet) and performance is based on prediction accuracy, convergence time, adversarial robustness, and privacy leakage resistance. The QIFL framework demonstrates superior performance over standard FL frameworks, achieving a prediction accuracy of97.8%(±0.3%)97.8 \% (\pm 0.3 \%)97.8%(±0.3%), a convergence time of52\mathbf{5 2}52percent quicker, a41%\mathbf{4 1 \%}41%lower adversarial success rate, and a reduced communication cost by 42%. This work reveals the effectiveness of quantum inspired computing and federated learning methods to produce an accurate, secure, and computationally efficient predictive healthcare model for the elderly population. The authors will continue to explore future work using Quantum Neural Networks (QNN) to enhance predictive intelligence and real-time diagnostics.