Recently, the broadcast of the TV drama 'I am a Criminal Investigation Officer' sparked a heated discussion among viewers, and the plot of using DNA testing technology to solve the case was even more touching. How can inspectors find evidence more quickly and accurately using the "tiny" biological samples discovered by police investigation? In order to solve the problem of rapid and accurate detection of trace biological samples, the research team of Shanghai University of Technology has developed the "AI fluorescence imaging - unfiltered fluorescence microscopy imaging technology" with the core concept of "AI for Science", making biological sample detection more efficient. This achievement was completed by the ultra precision optical manufacturing team led by Zhuang Songlin, academician of the CAE Member, dean of the School of Optoelectronic Information and Computer Engineering of Shanghai University of Technology, and Zhang Dawei, executive vice president of Shanghai University of Technology, in cooperation with Punan Branch of Renji Hospital affiliated to Shanghai Jiaotong University and Duke University in the United States, and can be widely used in biological diagnosis, clinical medicine, environmental monitoring, criminal investigation and exploration and other fields. The related paper titled "Filter free Fluorescence Microscopy Supporting Deep Learning" was recently published in the international academic journal "Science" sub issue "Scientific Progress". The original intention of the research on AI empowering optical devices is not for criminal investigation applications, but stems from the demand for clinical medical intelligence. Three years ago, relying on the medical engineering cross project of Shanghai University of Technology, Professor Dai Bo's team and Professor Guan Yangtai's team from Renji Hospital affiliated with Shanghai Jiao Tong University carried out a series of scientific research collaborations. During the cooperation process, both parties realized that advanced intelligent optoelectronic detection technology can provide comprehensive, accurate, and convenient means for clinical diagnosis, thereby benefiting a large number of patients. The unfiltered fluorescence microscopy imaging technology originated from this project. As a universal biomedical detection technology, it can fundamentally solve the pain points of traditional fluorescence microscopy imaging. Fluorescence imaging systems have been widely used in medical basic research and clinical practice, such as antibody detection for neuroimmune diseases, auxiliary detection and localization of tumor cells, monitoring of drug distribution and metabolism in the body, and so on Wang Kan, the research assistant of Guan Yangtai team, said that the existing fluorescence imaging system requires multiple sets of filtering components such as dichroic mirrors and filters to meet the requirements of multi band fluorescence imaging. In addition, when performing multi-channel fluorescence imaging, operators need to use mechanical devices to switch the fluorescence filtering components, which is not only time-consuming and cumbersome to operate, but also increases the complexity, volume, and cost of the imaging system. The research of the Dai Bo team on "AI empowered optical devices" can precisely solve this pain point. The team boldly proposed whether AI technology can replace traditional optical filtering components to locate and quantitatively analyze the specificity of fluorescence signals, achieving efficient and accurate detection and analysis of biological samples. The research on "digital virtual filters" launched from this has become the key to solving difficult problems. After efficiently restoring various fluorescent signals and identifying the research direction, new challenges have been presented to the research team. This research involves interdisciplinary integration and requires scholars from different fields to tackle it together. Previously, the research team gathered experts from fields such as optical engineering, artificial intelligence, and clinical medicine, forming a unique interdisciplinary advantage to facilitate the final implementation of the project. The team started from AI technology and developed a digital virtual filter, and proposed unfiltered fluorescence microscopy imaging technology. The fluorescence microscopy imaging system using this technology does not require expensive fluorescence filtering components and can reduce background noise through dark field illumination. After the imaging system obtains the image, it automatically selects the fluorescence channel through a neural network to accurately predict the fluorescence signal. The results indicate that the unfiltered fluorescence microscopy imaging system has good robustness and can efficiently and accurately restore fluorescence signals for different microscopic magnifications, fluorescent dye concentrations, and sample types, achieving high-sensitivity and high specificity fluorescence imaging. The first author of the paper, Dai Bo, gave an analogy: "Cells in a microscope are like starlight. We need to use different lenses to recognize different colors. If we want to recognize all these light points, we have to constantly change the lenses, which is time-consuming and laborious. Our research team has overturned this concept using AI technology. No matter how many different colored light points there are in the sky, they can all be clearly and quickly recognized with a regular telescope." In the study, the team conducted fluorescence imaging experiments on multi-color fluorescent quantum dot nanoparticles, cells co stained with multiple fluorescent dyes, tissue slices, and dynamic cells. In addition, they also used an unfiltered fluorescence microscopy imaging system for fibroblast activation protein expression analysis, detection of human esophageal tissue, human liver tissue slices, and a series of biological studies, as well as clinical testing experiments. In the past, pathological examination of a tumor slice required at least 20 minutes, but with this new technology, it only takes 4 minutes, increasing efficiency by as much as 5 times. New technologies will provide more assistance for doctors to accurately diagnose and protect the health of patients. The proposed unfiltered fluorescence imaging technology has been preliminarily validated in fluorescence microscopy imaging systems. This technology has enormous research value and application potential, and needs to be further transplanted into various fluorescence detection related instruments, such as confocal microscopy, fluorescence flow cytometry, etc., to promote the intelligent upgrading and replacement of existing biochemical detection instruments Zhang Dawei said. (New Society)