From the Clinic, to the Clinic: Improving the Fluorescent Imaging Quality of ICG via Amphiphilic NIR-IIa AIE Probe

Fluorescence imaging is crucial for providing detailed information in clinical practice. However, traditional first near-infrared (NIR-I) dyes such as indocyanine green (ICG) exhibit limitations such as shallow penetration depth, low contrast, and suboptimal clarity due to light scattering and autofluorescence. To overcome these drawbacks, we utilized a novel amphiphilic second near-infrared (NIR-II) aggregation-induced emission (AIE) probe (TCP) with an emission range beyond 1300 nm (NIR-IIa). Using approximately 200 co-registered NIR-I/NIR-IIa image pairs acquired with TCP, we trained a SwinUnet-based deep learning model to transform low-quality NIR-I ICG images into high-resolution NIR-IIa-like images. Owing to its superior brightness and photostability, TCP enhances in vivo fluorescent angiography, offering clearer vascular details and a higher signal-to-background ratio (SBR) in the NIR-IIa region, 2.6-fold higher than that of ICG in the NIR-I region. The deep learning model successfully converted blurred NIR-I images into high-SBR NIR-IIa-like images, achieving rapid imaging speeds without compromising quality. This work introduces a synergistic "probe-plus-AI" paradigm that substantially improves both the quality and speed of clinical fluorescence imaging, providing a pathway that is immediately translatable to enhanced diagnostics and image-guided surgery.