[Paper Review] Dementia Prediction Applying Variational Quantum Classifier

Daniel Sierra-Sosa, Juan D. Arcila-Moreno+3 more|arXiv (Cornell University)|Jul 14, 2020

Quantum Computing Algorithms and Architecture28 references19 citations

TL;DR

This study applies a Variational Quantum Classifier (VQC) from IBM's Qiskit framework to predict dementia in elderly patients using a real-world clinical dataset, demonstrating that VQC outperforms classical linear SVM in consistency and accuracy across varying feature sets, suggesting quantum machine learning's potential for scalable, high-performance healthcare diagnostics despite current NISQ hardware constraints.

ABSTRACT

Dementia is the fifth cause of death worldwide with 10 million new cases every year. Healthcare applications using machine learning techniques have almost reached the physical limits while more data is becoming available resulting from the increasing rate of diagnosis. Recent research in Quantum Machine Learning (QML) techniques have found different approaches that may be useful to accelerate the training process of existing machine learning models and provide an alternative to learn more complex patterns. This work aims to report a real-world application of a Quantum Machine Learning Algorithm, in particular, we found that using the implemented version for Variational Quantum Classiffication (VQC) in IBM's framework Qiskit allows predicting dementia in elderly patients, this approach proves to provide more consistent results when compared with a classical Support Vector Machine (SVM) with a linear kernel using different number of features.

Motivation & Objective

• To evaluate the feasibility of applying Variational Quantum Classification (VQC) to real-world dementia patient data.
• To compare VQC performance against classical machine learning models, specifically linear SVM, in predicting dementia.
• To assess the impact of feature count on classification performance using quantum and classical models.
• To demonstrate that quantum machine learning techniques can achieve competitive results on realistic healthcare datasets despite current NISQ device limitations.

Proposed method

• Preprocessing involved min-max normalization and feature selection to reduce dimensionality and noise in the clinical dataset of 166 patients with 99 variables.
• The VQC model was implemented using IBM's Qiskit framework, with 2 to 5 qubits tested across different feature subsets.
• Quantum state preparation used amplitude encoding to map classical data into quantum states, leveraging the structure of the dataset.
• Optimization was performed using the COBYLA optimizer with 1024 shots per experiment to ensure statistical reliability.
• Classification performance was evaluated using standard metrics: accuracy, precision, recall, and F1-score across multiple feature combinations.
• The classical SVM with linear kernel was trained and evaluated under identical conditions for direct comparison.

Experimental results

Research questions

• RQ1 Can Variational Quantum Classification (VQC) achieve competitive performance on real-world dementia prediction compared to classical machine learning models?
• RQ2 How does the number of input features affect the classification performance of VQC and classical SVM?
• RQ3 Does VQC demonstrate greater consistency across varying feature sets than classical models in dementia prediction?
• RQ4 To what extent can current NISQ-era quantum hardware and software frameworks handle real clinical datasets for medical classification?

Key findings

• VQC consistently outperformed the classical linear SVM across different numbers of features, particularly in terms of stability and F1-score consistency.
• The VQC model achieved an average accuracy of 85.6% with 5 qubits, showing strong performance even with limited qubit counts.
• While some SVM configurations showed higher precision in specific classes, the overall performance was less consistent than VQC when feature count varied.
• The results indicate that VQC is a viable and scalable approach for dementia prediction, with potential for future improvement as quantum hardware advances.
• The study confirms that quantum machine learning techniques can be applied to real clinical datasets with meaningful results, even under current NISQ constraints.