Benchmarking MedMNIST dataset on real quantum hardware

TL;DR

This paper addresses the problem of evaluating purely quantum models for medical image classification on real hardware, using MedMNIST datasets and a 127-qubit IBM Cleveland device. The authors implement a four-stage QML workflow—downsampling with angle encoding, device-aware circuit generation via the Élivágar framework, classical training with gradient-based optimization, and hardware-inference enhanced by dynamical decoupling, gate twirling, and M3 mitigation—yielding measurable ACC and AUC performance. Key contributions include the first hardware benchmark of MedMNIST with quantum-only models, an ablation study of error suppression/mitigation techniques, and a detailed comparison to classical baselines across multiple datasets, highlighting both the current limitations and potential quantum advantages in low-feature regimes. The results establish a practical benchmark for QML in healthcare, illustrating feasible quantum-classical boundaries on NISQ devices and guiding future improvements in circuit design and noise mitigation for real-world applicability.

Abstract

Quantum machine learning (QML) has emerged as a promising domain to leverage the computational capabilities of quantum systems to solve complex classification tasks. In this work, we present the first comprehensive QML study by benchmarking the MedMNIST-a diverse collection of medical imaging datasets on a 127-qubit real IBM quantum hardware, to evaluate the feasibility and performance of quantum models (without any classical neural networks) in practical applications. This study explores recent advancements in quantum computing such as device-aware quantum circuits, error suppression, and mitigation for medical image classification. Our methodology is comprised of three stages: preprocessing, generation of noise-resilient and hardware-efficient quantum circuits, optimizing/training of quantum circuits on classical hardware, and inference on real IBM quantum hardware. Firstly, we process all input images in the preprocessing stage to reduce the spatial dimension due to quantum hardware limitations. We generate hardware-efficient quantum circuits using backend properties expressible to learn complex patterns for medical image classification. After classical optimization of QML models, we perform inference on real quantum hardware. We also incorporate advanced error suppression and mitigation techniques in our QML workflow, including dynamical decoupling (DD), gate twirling, and matrix-free measurement mitigation (M3) to mitigate the effects of noise and improve classification performance. The experimental results showcase the potential of quantum computing for medical imaging and establish a benchmark for future advancements in QML applied to healthcare.

Figures (2)

• Figure 1: QML framework for medical image classification.
• Figure 2: MedMNIST datasets for 2D biomedical image classification.