Deep Learning Approach for Quantification of Fluorescently Labeled Blood Cells in Danio rerio (Zebrafish)

Abstract

Neutrophils are a type of white blood cell essential for the function of the innate immune system. To elucidate mechanisms of neutrophil biology, many studies are performed in vertebrate animal model systems. In Danio rerio (zebrafish), in vivo imaging of neutrophils is possible due to transgenic strains that possess fluorescently labeled leukocytes. However, due to the relative abundance of neutrophils, the counting process is laborious and subjective. In this article, we propose the use of a custom trained “you only look once” (YOLO) machine learning algorithm to automate the identification and counting of fluorescently labeled neutrophils in zebrafish. Using this model, we found the correlation coefficient between human counting and the model equals r = 0.8207 with an 8.65% percent error, while variation among human counters was 5% to 12%. Importantly, the model was able to correctly validate results of a previously published article that quantitated neutrophils manually. While the accuracy can be further improved, this model notably achieves these results in mere minutes compared with hours via standard manual counting protocols and can be performed by anyone with basic programming knowledge. It further supports the use of deep learning models for high throughput analysis of fluorescently labeled blood cells in the zebrafish model system.

Automated, high-throughput quantification of EGFP-expressing neutrophils in zebrafish by machine learning and a highly-parallelized microscope

“…A frequent question in digital cell counting is how a protocol rationalizes occluded cells or groups of cells with indistinguishable borders. Implementations can be found where researchers have used more advanced computational methods and counting strategies such as youonly-look-once (YOLO) convolutional networks [25,26,63], outlier rejection based on cell size or fluorescence intensity and watershed segmentation to define cell borders [64][65][66] which can yield improvements and high accuracy in terms of resolving individual cells. In our case, making use of the high parallelization of our imaging, we only require an estimation of cell count over a population and not an absolute value on a per fish basis and so further counting methods have not been found necessary.…”

“…Previously published algorithms to automatically quantify fluorescent immune cells in zebrafish [25,26] relied on time-consuming positioning and imaging one larva at a time, which dramatically reduced the throughput and scalability of the assay. Improvements in technology have recently allowed for the automation of this process using scanning microscopes [27,28] which can increase the number of fish that can be feasibly examined, however restrictions still exist with these methods with regard to the time required to capture and process data as well as requisite orientation and pigmentation level of the fish, again limiting utility.…”

Automated, high-throughput quantification of EGFP-expressing neutrophils in zebrafish by machine learning and a highly-parallelized microscope

“…A frequent question in digital cell counting is how a protocol rationalizes occluded cells or groups of cells with indistinguishable borders. Implementations can be found where researchers have used more advanced computational methods and counting strategies such as you-only-look-once (YOLO) convolutional networks [25,26,55], outlier rejection based on cell size or fluorescence intensity and watershed segmentation to define cell borders [56][57][58] which can yield improvements and high accuracy in terms of resolving individual cells. In our case, making use of the high parallelization of our imaging, we only require an estimation of cell count over a population and not an absolute value on a per fish basis and so further counting methods have not been found necessary.…”