Q RadFusion: Hybrid Quantum Classical Radiogenomic Framework for Breast Cancer Diagnosis

Area of Science:

• Radiogenomics
• Quantum Computing
• Biomedical Imaging

Background:

• Breast cancer is a leading global cancer in women, necessitating early and accurate diagnosis for improved survival.
• Radiogenomics offers precision diagnostics by integrating imaging phenotypes with genomic data.
• Classical machine learning models face challenges with high-dimensional, heterogeneous multimodal data in radiogenomics.

Purpose of the Study:

• To present Q RadFusion, a novel hybrid quantum-classical framework for enhanced breast cancer diagnosis.
• To fuse mammography and genomics data for improved diagnostic accuracy and reproducibility.
• To address limitations of classical models in handling complex multimodal data.

Main Methods:

• Implementation of Q RadFusion on CBIS-DDSM (mammography) and TCGA-BRCA (genomics) datasets.
• Utilized Quantum Approximate Optimization Algorithm (QAOA) for feature selection and Variational Quantum Circuits (VQC) for feature mapping.
• Employed ResNet for mammography features and Transformer for genomic features, followed by multimodal fusion.

Main Results:

• Q RadFusion achieved an Area Under the Curve (AUC) of 0.96 and 94% accuracy, outperforming baseline models.
• Ablation studies confirmed the efficacy of quantum components, with optimal performance at specific circuit parameters (L=6, Q=10, p=3).
• Demonstrated improved calibration and a significant reduction (~80%) in parameters compared to deep fusion networks.

Conclusions:

• Hybrid quantum-classical radiogenomic integration, as exemplified by Q RadFusion, offers accurate and reproducible diagnostic support for breast cancer.
• The framework shows strong potential for clinical translation in precision diagnostics.
• Q RadFusion advances the field of AI in medical imaging and genomics for oncology.