FDTD-Based Synthetic Dataset Generation and Quantum Ensemble Learning for Enhanced Microwave Tumor Localization

<title>Abstract</title> <p>Microwave-based tumor localization is a new non-invasive diagnostic tool that uses the differences in dielectric properties between healthy and cancerous tissues. Tumor tissue has a high water content, which leads to notable reflections of electromagnetic waves at the boundary between tissue types. This makes microwave imaging an effective method for detection. However, creating robust classification algorithms is often challenging due to the scarcity of labeled clinical datasets. This research presents a framework for generating high-quality synthetic microwave scattering data through Finite-Difference Time-Domain (FDTD) simulations. Our approach employs a multi-static MIMO setup and a differential imaging technique to separate tumor-specific signals from complex biological backgrounds. We assess the performance of a traditional Support Vector Machine (SVM) alongside individual quantum models, specifically the Quantum Multi-Encoder Fusion Classifier (QMEFC) and the Hybrid Quantum Spatial Residual Network (HQ-SRN). To enhance diagnostic accuracy, this research introduces the Quantum-Powered Fusion Classifier Ensemble (QPFC-E), which integrates these models via Dempster-Shafer (DS) belief fusion and AUC-weighted trust parameters. Experimental results show that the QPFC-E achieves an overall accuracy of 87.00% and a Macro AUC-ROC of 0.9354. Notably, the ensemble method achieves a high recall of 61.3% for difficult tumor types, a substantial improvement over the classical baseline. This work demonstrates the potential of hybrid FDTD-QML frameworks in improving the accuracy of microwave diagnostics.</p>