Deep-Learning-Enhanced Bioimaging Via Energy Traps Regulated Lanthanide Nanoparticles

High-resolution biological imaging in deep tissues holds substantial importance for advancing precision medicine. Currently, the lanthanide-doped nanoparticles enable in vivo near-infrared imaging but face a critical trade-off: Er^{3 +}-based emission at around 1530 nm enables high optical resolution yet suffers from limited tissue penetration due to water absorption, whereas nanoprobes emitting at approximately 980/1060 nm offer deeper penetration with high brightness but compromised resolution due to higher tissue scattering at shorter wavelength. This fundamental contradiction between imaging depth and resolution remains a key challenge. Herein, we introduce the concept of energy traps to actively regulate energy distribution within lanthanide nanoparticles via excitation-wavelength switching and directional energy transfer modulation. This strategy enables controlled access to either sensitizers (Yb³⁺, Nd³⁺) self-emission or efficient sensitization of activators (Er³⁺), thereby allowing selective enhancement of emission channels. By integrating the deep-tissue penetration capability of short-wavelength sensitizers (980 and 1060 nm) with the high-resolution emission of long-wavelength activators (1530 nm) through a deep-learning-based network, we successfully achieved a 93% enhancement in imaging performance for short-wavelength probes, offering a robust and adaptable platform for high-contrast deep-tissue bioimaging and future point-of-care diagnostics.