Our team has developed an AI-powered varifocal meta-optical endomicroscopy system that integrates a deep learning-based HiLo sectioning algorithm, a telecentric design, and a varifocal metalens. Using a single input image, the system can reconstruct high-contrast optical sectioning images in real time, achieving a 50-fold improvement in processing efficiency. It has been successfully applied to in vivo microvascular imaging, demonstrating high resolution and real-time performance, and is suitable for medical applications such as neural imaging and brain surgery.

This system integrates a rotatable Moiré metalens with AI deep learning algorithms to achieve high-contrast, multi-depth (multi-plane), real-time 3D optical sectioning. The metalens, combined with a telecentric design, forms a compact structure that stably provides high-quality images across different depths. When combined with an RCNN model, the system can reconstruct HiLo section images from a single input, enhancing processing efficiency by up to 50-fold. The system has been successfully applied to in vivo cerebral microvascular imaging, demonstrating high resolution, low latency, and strong potential for clinical applications.

Developed the “AI-powered Varifocal Meta-optics Endomicroscopy System,” which offers high-resolution, real-time imaging with minimal invasiveness, making it suitable for neurosurgical, cerebral vascular observation in high-precision clinical settings. Compared to confocal and multiphoton endoscopes, this system excels in compactness, imaging speed, and operational simplicity. With AI-based image reconstruction and a modular optical design, it holds strong potential for commercialization.