Extracting Patterns of Lymphocyte Fluorescence from DigitalMicroscope

A system for full-automatic extraction of uorescence information from lymphocytes in digital microscope images is presented. It is applied on samples of lymphocytes invading muscle tissue. From a single sample images are taken using n diierent uorochrome antibody markers. In every image a diierent subset of the lymphocytes appears through uorescence. The uorescent cells in every image are detected by an artiicial neural network. After detection, corresponding uorescences of each cell in diierent images are collected in lists of length n, called marker combination patterns. Each binary element of a pattern represents, if a cell was uorescent (1), or not (0) in a particular image. The whole algorithm of detection and correspondence analysis is easy to adapt and computes fast. This automation allows us to gain reproducible data and opens the door for a statistical analysis of a large number of uorescence patterns.