From Quantum Metrics to Classical Intelligence: Deep Learning Tc-99m Stability and Decay with Sub-Femtosecond Accuracy

We present a novel deep learning framework that achieves quantum-level precision in modeling the nuclear properties of Technetium-99m (Tc-99m), a cornerstone isotope in diagnostic nuclear medicine. By encoding quantum information measures—such as Rényi-2 entropy and purity—alongside Nikiforov-Uvarov (NU) functionals into structured input tensors, our classical deep neural network (DNN) effectively learns complex relationships traditionally reserved for quantum or hybrid quantum-classical systems. The model uncovers three key findings: (1) a 660-attosecond coherence-based stability threshold, (2) a strong inverse correlation between quantum purity and decay stability (Pearson ρ = − 0.76 ± 0.05), and (3) optimal entropy windows between 0.5 and 1.5 nats governing Tc-99m decay modes. Compared to conventional nuclear decay models, our approach achieves a 28–32% improvement in half-life prediction accuracy, with performance validated against the IAEA nuclear decay database. These results demonstrate that carefully engineered quantum-informed features can endow classical AI with the capacity to emulate quantum behaviors—without requiring quantum hardware. This work establishes a scalable blueprint for deploying quantum-informed classical machine learning in radiopharmaceutical development, nuclear diagnostics, and beyond.