A model for generating paired complete blood count and oblique back-illumination capillaroscopy data in tissue-realistic microfluidic chambers.

Oblique back-illumination capillaroscopy (OBC) has recently demonstrated clear images of unlabeled human blood cells in vivo. Combined with deep learning-based algorithms, this technology may enable non-invasive blood cell counting and analysis as flowing red blood cells, platelets, and white blood cells can be observed in their native environment. To harness the full potential of OBC, new techniques and methods must be developed that provide ground truth data using human blood cells. Here we present such a model, where human blood cells with paired ground truth information are imaged flowing in a custom tissue-mimicking microfluidic device. This model enables the acquisition of OBC datasets that will help with both training and validating machine learning models for applications including the complete blood count, specific blood cell classification, and the study of hematologic disorders such as anemia.