Integrated quantum-classical hybrid architectures for robust lung lesion segmentation in volumetric CT video data samples

Segmentation of lung lesions in volumetric CT data is crucial for the clinical aspects of diagnosis, therapy planning, and monitoring disease progression. Currently, deep learning applications are unable to model spatiotemporal coherency alongside anatomical consistency and uncertainty-aware refinement across sequential slices. In this study, we propose a hybrid quantum–classical framework that would accommodate multiple innovative modules. The architecture features a Quantum Latent Entanglement Consistency validator to establish spatiotemporal coherence across slices by maximizing von Neumann entropy. A Quantum-Classical Interventional Gradient Alignment ensures the harmony of gradients between classical CNN encoders and quantum discriminators. Further, the Temporal Quantum Attention for Boundary Stabilization captures the temporal context in the boundary refinement using controlled quantum gates. Alongside these, a Quantum-Enhanced Structural Similarity Feedback mechanism is proposed that exploits anatomical priors for retrofitting spatial lesion structures, as well as a Hybrid Quantum Adversarial Ensemble Validation, which provides confidence-aware validity through disagreement modeling. Collection and experimental evaluations over LIDC IDRI, NSCLC-Radiomics, and MosMedData datasets depict that the entirety of the systems significantly increases the Dice Similarity Coefficient by 5–7%, holds Hausdorff Distance lower at 10–12%, narrows down the over-segmentation errors by 8–10%, while reducing overall false positives near lung boundaries by 15% or even less. This represents a significant advancement toward fusing quantum learning with clinical-grade imaging pipelines, demonstrating clear improvements in segmentation stability, precision, and trustworthiness in real-world settings.