Scholarlink

<p>In this study, we propose a quantum variational autoencoder (QVAE)-based framework for compressing and classifying cancer images in a quantum machine learning context. Our method addresses the space complexity limitations faced by quantum systems-particularly in the Noisy Intermediate-Scale Quantum (NISQ) era-by reducing the dimensionality of input features before classification. We evaluate our approach on three publicly available cancer image datasets: histopathologic scans, blood smear images, and dermatoscopic skin lesions (HAM10000). Three types of image features are explored: raw images, thresholded regions of interest (ROIs), and contour-based extractions. Experimental results show that QVAE achieves high compression fidelity (up to 0.998) while preserving classification accuracy within an acceptable margin. When using compressed features, our best-performing model achieves up to 98.0% classification accuracy. These findings demonstrate that QVAE can provide significant input reduction with minimal performance degradation, enabling more scalable quantum-classical hybrid pipelines for medical image analysis.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Quantum Variational Autoencoder for Feature Compression and Classification of Cancer Images

  • 期刊论文

  • 发表日期：

    2026年1月30日

查看全文
  • M. Bagus Andra,
  • Vicky Zilvan,
  • R. Sandra Yuwana,
  • Dikdik Krisnandi,
  • Endang Suryawati,
  • Ana Heryana,
  • Hilman F. Pardede

摘要

In this study, we propose a quantum variational autoencoder (QVAE)-based framework for compressing and classifying cancer images in a quantum machine learning context. Our method addresses the space complexity limitations faced by quantum systems-particularly in the Noisy Intermediate-Scale Quantum (NISQ) era-by reducing the dimensionality of input features before classification. We evaluate our approach on three publicly available cancer image datasets: histopathologic scans, blood smear images, and dermatoscopic skin lesions (HAM10000). Three types of image features are explored: raw images, thresholded regions of interest (ROIs), and contour-based extractions. Experimental results show that QVAE achieves high compression fidelity (up to 0.998) while preserving classification accuracy within an acceptable margin. When using compressed features, our best-performing model achieves up to 98.0% classification accuracy. These findings demonstrate that QVAE can provide significant input reduction with minimal performance degradation, enabling more scalable quantum-classical hybrid pipelines for medical image analysis.