Adversarial Feature Fusion in Hybrid Quantum-Classical Models for Endoscopic Image Classification

• Grant number: 26/11390-4
• Support Opportunities: Scholarships abroad - Research Internship - Scientific Initiation
• Start date: July 01, 2026
• End date: August 02, 2026
• Field of knowledge: Physical Sciences and Mathematics - Computer Science - Computer Systems
• Principal Investigator: João Paulo Papa
• Grantee: Yasmin Rodrigues Sobrinho
• Supervisor: Christoph Palm
• Host Institution: Faculdade de Ciências (FC). Universidade Estadual Paulista (UNESP). Campus de Bauru. Bauru , SP, Brazil
• Institution abroad: Ostbayerische Technische Hochschule Regensburg (OTH Regensburg), Germany
• Associated to the scholarship: 24/00117-0 - Quantum Convolutional Neural Network for Breast Cancer Detection through Mammograms, BP.IC

Abstract

This project proposes hybrid architectures that combine quantum and classical machine learning techniques to address the complex challenges of endoscopic medical image analysis. The research focuses on feature representations extracted from quantum and classical convolutional neural networks, aiming to leverage their complementary strengths for clinical classification tasks. In particular, the project introduces the Hybrid Classical-Quantum Adversarial Feature Fusion (HCQ-AFF) strategy. In this framework, a frozen classical network provides a stable reference, while a discriminator network actively encourages the quantum model to learn more informative and discriminative representations. This adversarial mechanism promotes robust feature alignment and more effective optimization of the quantum component. Furthermore, the proposal explicitly accounts for the practical limitations of current quantum hardware, such as limited circuit depth, thereby ensuring algorithmic compatibility with near-term intermediate-scale quantum devices. The objective is to establish a robust methodological framework for the deep integration of quantum and classical representations, thereby supporting accurate and reliable clinical decision-making based on gastrointestinal endoscopic imaging. (AU)