Neural Networks for Medical Data Processing: Segmentation of Fluorescence Lifetime Imaging Microscopy

Fluorescence lifetime imaging microscopy (FLIM) is a powerful tool for visualizing the distribution of fluorescence lifetimes of fluorophores within a tissue, and is used for optical metabolic imaging. However, processing the large volumes of data generated by FLIM can be time-consuming and challenging for an untrained operator. Convolutional neural networks (CNNs) have demonstrated its effectiveness for tissue image and FLIM segmentation, making them a promising tool for automated analysis of FLIM images. The article presents a method for segmenting FLIM images of liver tissue during the regeneration using a CNN based on the UNet++ architecture. The CNN was trained to detect cell boundaries and nuclei, and the resulting masks were used to approximate the fluorescence decay curves for the FLIM-image data. The results show that the CNN is effective in segmenting FLIM images of cells and tissues, and obtaining fluorescence decay curves. The use of CNNs and other machine learning methods in medical image segmentation and analysis will improve and speed up diagnosis and reduce diagnostic errors.