Laser & Optoelectronics Progress, Volume. 62, Issue 18, 1817022(2025)

Deep Learning-Based Resolution Enhancement Method for NIR-II Fluorescence Imaging (Invited)

Shiyi Peng1, Yuhuang Zhang1, Xiaolong Liu2, Xiaoxiao Fan2, Hui Lin2, and Jun Qian1、*
Author Affiliations
  • 1Centre for Optical and Electromagnetic Research, College of Optical Science and Engineering, International Research Center for Advanced Photonics, Zhejiang University, Hangzhou 310058, Zhejiang , China
  • 2Department of General Surgery, Sir Run Run Shaw Hospital, School of Medicine, Zhejiang University, Hangzhou 310058, Zhejiang , China
  • show less
    Figures & Tables(6)
    NIR-II imaging system. (a) Schematic diagram of NIR-II fluorescence imaging system; (b) comparison of pixel size between NIR-II detector and visible light camera, and its impact on imaging quality, enlarged area is imaging effect of fine blood vessels under resolution constraint; (c) schematic diagram of Real-ESRGAN network architecture
    Performance comparison of different super-resolution methods in NIR-II mouse blood vascular imaging. (a) Evaluation process of model performance; (b) comparison of reconstruction results for mouse back blood vessels by different methods, lower areas are magnified details; (c) LPIPS and PIQE of different methods based on 160 test images
    Application of Real-ESRGAN in NIR-Ⅱc band imaging. (a) Original NIR-Ⅱc images and processing workflow in different scenarios; (b) high-resolution reconstruction results processed by different methods, lower area is magnified details, scale bar is 5 mm; (c) intensity distributions of multiple blood vessels
    Performance of Real-ESRGAN in clinical application of diabetic foot. (a) Clinical data acquisition process; (b) high-resolution foot blood vascular images processed by different methods, right area shows magnified details; (c) intensity distribution of multiple blood vessels
    • Table 1. FWHM values for the blood vessels indicated in Fig. 3(c) determined by multi-peak fitting

      View table

      Table 1. FWHM values for the blood vessels indicated in Fig. 3(c) determined by multi-peak fitting

      RegionNumberBilinearBicubicMSRResNetRRDBNetReal-ESRGAN
      Abdomen10.2970.2870.3340.2720.193
      20.2510.2310.2880.2360.150
      30.7700.7210.8900.9400.821
      Intestine10.2310.2270.1590.2450.114
      20.2160.2020.2290.2220.175
      30.4740.3650.3080.2930.197
      40.2450.2320.2740.2710.209
      Leg10.7350.7290.7430.6190.382
      20.2060.2050.2130.2170.202
      30.7580.7300.7870.7540.610
    • Table 2. FWHM values for the blood vessels indicated in Fig. 4(c) determined by multi-peak fitting

      View table

      Table 2. FWHM values for the blood vessels indicated in Fig. 4(c) determined by multi-peak fitting

      Vessel numberBilinearBicubicMSRResNetRRDBNetReal-ESRGAN
      154.854.755.056.049.6
      247.948.046.745.637.4
      328.527.930.230.721.6
      450.149.851.051.046.2
      565.265.764.764.663.5
    Tools

    Get Citation

    Copy Citation Text

    Shiyi Peng, Yuhuang Zhang, Xiaolong Liu, Xiaoxiao Fan, Hui Lin, Jun Qian. Deep Learning-Based Resolution Enhancement Method for NIR-II Fluorescence Imaging (Invited)[J]. Laser & Optoelectronics Progress, 2025, 62(18): 1817022

    Download Citation

    EndNote(RIS)BibTexPlain Text
    Save article for my favorites
    Paper Information

    Category: Medical Optics and Biotechnology

    Received: May. 13, 2025

    Accepted: Jun. 17, 2025

    Published Online: Sep. 9, 2025

    The Author Email: Jun Qian (qianjun@zju.edu.cn)

    DOI:10.3788/LOP251220

    CSTR:32186.14.LOP251220

    Topics