Recently, the research group of Professor Zhou Pinghong of the Endoscopy Center of Zhongshan Hospital of Fudan University cooperated with the team of Professor Ji Minbiao of the Department of Physics of Fudan University to develop a fast pathological imaging technology based on femtosecond stimulated Raman technology, and successfully achieved Accurate real-time intelligent diagnosis of endoscopic biopsy tissue. The relevant research results are online under the title "Instant diagnosis of gastroscopic biopsy via deep-learned single-shot femtosecond stimulated Raman histology" ("Instant diagnosis of gastroscopic biopsy via deep-learned single-shot femtosecond stimulated Raman histology") Published in Nature Communications [Nature Communications, 13, 4050 (2022)]. Dr. Liu Zhijie and Ao Jianpeng from the Department of Physics of Fudan University, Dr. Su Wei, a doctoral student from Zhongshan Hospital of Fudan University are the co-first authors, Professor Ji Minbiao from the Department of Physics of Fudan University, Professor Zhou Pinghong and Hu Hao from the Endoscopy Center of Zhongshan Hospital of Fudan University The doctor is the co-corresponding author of and . This result is an important result of the smart medical care project of Shanghai Municipal Health and Family Planning Commission led by Professor Zhou Pinghong. It has also received funding from the key research and development projects of the Ministry of Science and Technology, the general project of the National Foundation of China, , and the Fudan University Medical-Industrial Integration Project. support.
Figure 1 (a) Illustration of stimulated Raman imaging of gastroscopic biopsy of tissue. (b) Conversion from a single femtosecond stimulated Raman scattering image to a pair of picosecond stimulated Raman scattering images using a U-shaped deep learning network.
In the pathological diagnosis process of conventional gastric endoscopic biopsy, the removed tissue needs to be embedded in paraffin , sectioned and stained. It takes several days to get the diagnostic result, and it is impossible to provide a timely diagnosis that matches the endoscopy. information. Stimulated Raman scattering (SRS) microscopy enables rapid, label-free molecular imaging of multi-component detection of biological tissues . Preliminary studies have shown that through dual-channel imaging of two biological macromolecules, proteins and lipids, histological information similar to traditional pathology (HE staining) can be obtained without any processing of biological tissue specimens, which is expected to provide surgical Real-time pathological diagnosis. However, current stimulated Raman imaging technology requires the use of picosecond laser pulses to achieve Raman spectrum resolution, combined with wavelength or pulse delay adjustment to achieve spectrum selection, which greatly limits the sensitivity and speed of imaging. On the other hand, femtosecond stimulated Raman can achieve high detection sensitivity and speed due to its higher laser peak power, but the broad-spectrum femtosecond pulse does not have spectral and molecular resolution capabilities and cannot be directly used for pathological imaging. . Therefore, how to parse and restore the multi-channel picosecond molecular image from the single-frame spectrum integrated femtosecond image is the key challenge and innovation of this project.
In order to overcome this technical problem, Professor Ji Minbiao's team took advantage of the inherent spectral-spatial correlation of biological tissues, combined with deep learning U-shaped network, and trained a large number of one-to-one corresponding femtosecond/picosecond tissue images to achieve from Perfect mapping of a single femtosecond image to a lipid/protein dual-channel picosecond image enables histopathological imaging based on femtosecond laser pulses (Figure 1). On this basis, Professor Zhou Pinghong's research group used deep learning femtosecond stimulated Raman imaging technology for rapid imaging, diagnostic classification and pathological segmentation of gastric cancer endoscopic biopsy tissue, achieving quasi-real-time pathology within 60 seconds. Diagnose, and reveal tumor heterogeneity within tissue (Video 1).
Video 1. Stimulated Raman images of endoscopic biopsy tissue reveal spatial heterogeneity of gastric cancer .
This research provides new solutions for simplifying complex stimulated Raman imaging systems, breaks through the speed limit of stimulated Raman pathology microscopes, expands the biomedical application scope of optical imaging technology, and provides endoscopic biopsy The rapid diagnosis of specimens provides a new method and a new technical basis for the clinical translation of stimulated Raman pathology microscopy.
paper link: https://doi.org/10.1038/s41467-022-31339-8
