# Real‐Time In Vivo Cellular‐Level Imaging During Puncture

## TL;DR

A new microendoscopy needle with AI enables real-time cellular imaging during biopsies, improving accuracy and reducing false negatives.

## Contribution

The AIM needle integrates adaptive optics and AI for real-time in vivo cellular imaging during puncture procedures.

## Key findings

- The AIM needle provides diffraction-limited imaging of layered microstructures in mouse organs and tumors.
- AI algorithms enable in situ pathology-like analysis and tumor/normal tissue discrimination during puncture.
- The device improves biopsy targeting accuracy and reduces false-negative rates in narrow lumens and microlesions.

## Abstract

In the pursuit of precise disease diagnosis, accurate tissue sampling during biopsy is critical. Current CT/ ultrasound‐guided biopsies provide macroscopic localization but lack real‐time cellular‐resolution visualization during puncture, particularly for deep, narrow lumens or small lesions, thereby increasing false‐negative and nondiagnostic sampling risks. It's desired to have a puncture with microscopy imaging. Here, we demonstrate an artificial‐intelligence‐empowered integrative‐light‐field microendoscopy (AIM) needle. This photonic‐mechanically co‐engineered probe overcomes dynamic diffraction limits via a closed‐loop adaptive optics system integrated in a 25G biopsy needle, enabling diffraction‐limited imaging during in vivo puncture. AIM needle resolved characteristic layered microstructures throughout mouse organs (parenchymal/hollow) and pulmonary tumors via synergistic light‐field modulation, leveraging K‐means and convolutional neural networks to enable in situ pathology‐like analysis and tumor/normal tissue discrimination along puncture paths. AIM needle demonstrates dual clinical potential as a complement to macroscopic guidance: potentially providing histology‐like feedback without interrupting procedures while enhancing biopsy targeting accuracy through navigation integration, reducing false‐negative rates in narrow lumens and microlesions, thereby improving early detection sensitivity.

We present an artificial‐intelligence‐empowered integrative‐light‐field microendoscopy (AIM) needle that delivers real‐time in vivo, diffraction‐limited cellular‐level imaging during puncture and visualizes layered microstructures along the needle path. As a microscopic complement to CT/ultrasound, it improves sampling localization and adds preliminary tumor/normal discrimination, potentially reducing false‐negative rates in biopsies of narrow lumens and microlesions, thereby improving early‐detection sensitivity.