A Novel algorithm for detecting isolated green fluorescently-labeled cells in Cryo-imaging data

Abstract

Cryo-imaging is a biomedical imaging technology for studying cellular biodistribution in small animal models. It can be used to locate cells of interest anywhere in a whole animal scale with single cell sensitivity. In this study, we are interested in green fluorescently labeled T-cells in the liver of the Graft versus-Host disease mouse model. However, the detection of green fluorescently labeled cells is quite highly challenging and difficult due to the autofluorescence in the liver, especially bile duct and gall bladder. They have the same spectrum as the signal of T-cells. We observed that autofluorescence mostly tended to form into the dense structure in 3D volume whereas the cell signals of interest were distributed and isolated throughout the liver tissue. We developed an algorithm that consisted of two essential parts: the T-cell signal detection and the removal of the autofluorescent signals. The detection part consisted of thresholding on imaging data converted by Mexican hat filtering and Top-hat transformation. The second part was measuring the voxel density in 3D space with Mean Inter-Particle Distance for eliminating noises. In this study, we used both synthetic data and real data to test the algorithm performance. We found that the sensitivity and specificity of detection were around 80-90% and 98%, respectively. In conclusion, we successfully developed an algorithm for detecting green fluorescently labeled cells and cleaning the structured autofluorescent signals for the first time. We believe that this research is a further development of the capability and efficiency of Cryo-imaging technology. As a result, the technology should become more well-known to the medical science community, and it will greatly benefit the development of small animal research in the future.