GBIQ: a non-arbitrary, non-biased method for quantification of fluorescent images.

Fluorescent cytometry is an indispensable method to obtain quantitative data from fluorescent staining. For tissue sections and substrate-attach cultured cells that should be analyzed in situ, a combination of fluorescent microscopy and image analysis is a standard solution for obtaining reliable quantitative data from them. A starting process of the image analysis would be identification of each cell by cell-segmentation using nuclear and/or cytoplasmic counterstaining, the process eventually defines quality of the outcome. Thus, once the segmentation was established, fluorescent intensities of other molecular entities in question within the cell are measured and stored at cellular basis for further phenotypic analysis1. The segmentation process, therefore, should be carried out with a great deal of attention to reach optimal cell-segmentation while eliminate unwanted artifacts. This means that there are a plenty of rooms of arbitration and bias at the starting process, and requirement of fine-tunings and human interventions on image-to-image basis during the process, which are potentially time-consuming, tedious and often leading to poor reproducibility of the outcomes. These drawbacks of the cell-segmentation method would also potentially hamper a streamlined implementation of the image measurement method with an automated high-throughput data analysis workflow. In fact, many cell-segmentation methods have been developed, none of them is applicable universally, algorithms and parameter-sets should be tailored for subject specific manner2. In our daily research activities at a bench, we often encounter a situation where we need quick quantitative assessment of fluorescent images that were produced by initial experiments. Even though cell-segmentation methods are robust and reliable, they are quite unsuitable in the situation due to methods’ demand for parameter settings. Especially, if the experimental subjects are densely packed cells including colonies of embryonic stem cells (ESCs) or condensing mesenchymal cells in tissue sections, it would be even more hard applying the cell-segmentation methods onto the subjects. Unlike mono-layered culturing cells, these subjects often contain complex cellular phenotypical heterogeneity and architectural diversity in a context of high-density of cells. This nature of subjects makes the cell-segmentation difficult without the tailored algorithm and a great deal of fine-tunings to the set of parameters. These arbitrary procedures inevitably take our labour and time, hence in this situation, a quick, easy and ideally parameter-free quantification method is in great demand. A solution to the problem would be a non-arbitrary and non-biased quantification method that processes fluorescent images under minimum set of rules. One possible way to attain this could be abandoning the cell-segmentation from the process. GBIQ (stands for Grid Based Image Quantification) is designed for reducing human intervenes as little as possible to achieve both reproducibility of data and streamlined workflow. Instead of cell identification by the cell-segmentation, GBIQ employs a tiling of fixed size grids and utilizes statistics within and among the grids to quantify fluorescent images. Although GBIQ does not identify single cells in principle, the method yet produces quantitative dataset that can be classified to sub-populations among which phenotypic features vary in a subject image dataset. Here we report the process of GBIQ in comparison to the cell-segmentation based method, verifying GBIQ is a practical alternative to the segmentation-based method. GBIQ particularly well performs in cases the cell density of subject is high, where the segmentation method would require elaborative fine-tunings of parameter-sets to identify each cell, as exampled here applying the both methods on same image datasets obtained from dense colonies of mouse ESCs (mESCs). We also apply GBIQ on various tissue sections from developing mouse embryos to elucidate a gene regulatory network.