AI-Powered Super-Resolution Imaging Enables Single-Shot, High-Fidelity Live-Cell Observation

A research team led by Professor Chao Zuo at Nanjing University of Science and Technology (NJUST) has introduced an AI-driven super-resolution imaging approach called eDL-cSIM, designed to overcome key limitations in current microscopy techniques.

Capturing the interactions of organelles, proteins, and molecular components within living cells is vital for understanding cellular function, disease mechanisms, and drug responses. Super-resolution fluorescence microscopy has greatly advanced researchers’ ability to see beyond the classical diffraction limit, revealing cellular structures in remarkable detail. However, these methods still face notable challenges: many require multiple sequential images or scanning, leading to longer imaging times, motion artifacts, and increased risk of phototoxicity and photobleaching due to higher photon doses. Some techniques also rely heavily on post-processing algorithms, which can introduce reconstruction artifacts, especially under low signal-to-noise conditions, limiting the reliability and broader application of super-resolution imaging, especially for live-cell studies that demand fast, gentle, and accurate observation.

To address these issues, the NJUST team developed eDL-cSIM. This technique uses a six-beam interference strategy, encoding super-resolution information from multiple orientations in a single camera exposure. As a result, both the photon dose and the number of required frames are reduced by more than nine times. The method also employs a custom ensemble neural network that integrates Transformer modules and multi-model fusion strategies to reconstruct images with 100 nm lateral resolution from just one frame. “Our method merges the strengths of structured illumination and deep learning to decode fine spatial information directly from minimally acquired data,” said Prof. Zuo, senior author of the study published in PhotoniX. “This not only speeds up the imaging process, but also minimizes phototoxicity, making it better suited for live-cell observation.”

In practical demonstrations, eDL-cSIM visualized the dynamic remodeling of intracellular structures, such as mitochondrial networks undergoing fission and fusion—processes essential for cellular health and energy regulation. Disruptions in these processes are linked to pathologies such as neurodegenerative diseases and metabolic syndromes. “By recording these processes at high resolution and frame rates, eDL-cSIM could help scientists better understand cell biology in health and disease,” added Professor Zuo.

Compared to traditional methods, eDL-cSIM offers superior imaging speed, reconstruction quality, and robustness across different sample types and cellular features, making it highly adaptable for live-cell imaging and tracking dynamic processes, Professor Zuo noted. Overall, eDL-cSIM exemplifies a new generation of intelligent microscopes, combining advanced optics with deep learning to enable faster, gentler, and more precise observation of subtle cellular dynamics.