Integrating deep learning with microfluidics for biophysical classification of sickle red blood cells adhered to laminin

Author summaryAmong the most serious consequences of sickle cell disease are enhanced adhesive intereactions between sickle red blood cells and the walls of blood vessels, leading to potentially fatal vaso-occlusive crises. To better understand these interactions, microfluidic experiments inject patient blood into cavities lined with proteins that tether the red blood cells. The numbers and shapes of the adhered cells provide important clues about the underlying biophysics of the disorder, and could allow detailed patient-specific monitoring of the disease progression. However a major bottleneck in moving this technology from the lab to the clinic has been reliance on specially trained workers to manually count and classify cells, which can take hours for each experimental image. Using deep learning image analysis, we demonstrate a way to address this problem: an automated system for biophysical characterization of adhered sickle red blood cells in microfluidic devices. Our system can achieve accuracies that not only match those of human experts, but are more consistent, showing less variance in counts on repeat trials. Moreover the processing takes minutes rather than hours per image. Our results are a first step toward enabling fast and affordable next-generation microfluidic monitoring tools for sickle cell disease. Sickle cell disease, a genetic disorder affecting a sizeable global demographic, manifests in sickle red blood cells (sRBCs) with altered shape and biomechanics. sRBCs show heightened adhesive interactions with inflamed endothelium, triggering painful vascular occlusion events. Numerous studies employ microfluidic-assay-based monitoring tools to quantify characteristics of adhered sRBCs from high resolution channel images. The current image analysis workflow relies on detailed morphological characterization and cell counting by a specially trained worker. This is time and labor intensive, and prone to user bias artifacts. Here we establish a morphology based classification scheme to identify two naturally arising sRBC subpopulations-deformable and non-deformable sRBCs-utilizing novel visual markers that link to underlying cell biomechanical properties and hold promise for clinically relevant insights. We then set up a standardized, reproducible, and fully automated image analysis workflow designed to carry out this classification. This relies on a two part deep neural network architecture that works in tandem for segmentation of channel images and classification of adhered cells into subtypes. Network training utilized an extensive data set of images generated by the SCD BioChip, a microfluidic assay which injects clinical whole blood samples into protein-functionalized microchannels, mimicking physiological conditions in the microvasculature. Here we carried out the assay with the sub-endothelial protein laminin. The machine learning approach segmented the resulting channel images with 99.1 +/- 0.3% mean IoU on the validation set across 5 k-folds, classified detected sRBCs with 96.0 +/- 0.3% mean accuracy on the validation set across 5 k-folds, and matched trained personnel in overall characterization of whole channel images with R-2 = 0.992, 0.987 and 0.834 for total, deformable and non-deformable sRBC counts respectively. Average analysis time per channel image was also improved by two orders of magnitude (similar to 2 minutes vs similar to 2-3 hours) over manual characterization. Finally, the network results show an order of magnitude less variance in counts on repeat trials than humans. This kind of standardization is a prerequisite for the viability of any diagnostic technology, making our system suitable for affordable and high throughput disease monitoring.

Automated summary

This study introduces an automated system for biophysical characterization of adhered sickle red blood cells in microfluidic devices using deep learning image analysis, showing that the system achieves accuracies matching human experts but with less variance in counts and faster processing time per image. This standardized and reproducible approach holds promise for affordable and high throughput disease monitoring.