Abstract
Imaging and characterizing the dynamics of cellular adhesion in blood samples is of fundamental importance in understanding biological function. In vitro microscopy methods are widely used for this task but typically require diluting the blood with a buffer to allow for transmission of light. However, whole blood provides crucial signaling cues that influence adhesion dynamics, which means that conventional approaches lack the full physiological complexity of living microvasculature. We can reliably image cell interactions in microfluidic channels during whole blood flow by motion blur microscopy (MBM) in vitro and automate image analysis using machine learning. MBM provides a low cost, easy to implement alternative to intravital microscopy, for rapid data generation where understanding cell interactions, adhesion, and motility is crucial. MBM is generalizable to studies of various diseases, including cancer, blood disorders, thrombosis, inflammatory and autoimmune diseases, as well as providing rich datasets for theoretical modeling of adhesion dynamics. Monitoring cellular adhesion in blood samples is of fundamental importance, but conventional approaches may not capture the full physiological complexity of microscale flow in living contexts. Here, the authors present motion blur microscopy and machine learning-based image analysis to study cell interactions in microfluidic channels during whole blood flow.
Introduction
Cell adhesion—the process by which leukocytes, platelets, or cancer cells adhere to blood vessel walls—is a fundamental driver of inflammation, immunity, and disease progression. To study these dynamics in a controlled environment, researchers frequently employ microfluidic devices. However, the inherent opacity of whole blood presents a significant challenge for standard microscopy, often necessitating sample dilution with buffers.
This dilution introduces a critical methodological flaw: it removes essential plasma proteins and eliminates the "crowding" effect of red blood cells (RBCs). In vivo, this crowding naturally pushes other cells toward the vessel walls, a process known as margination. While intravital microscopy (imaging within a living organism) preserves these physiological conditions, it is prohibitively expensive and technically demanding. Motion Blur Microscopy (MBM) offers a solution by providing "intravital-like" realism in a simple, benchtop in vitro setup.
Methodology: The Physics of Motion Blur
The core innovation of MBM lies in its counter-intuitive use of long exposure times—up to 1200 ms—combined with reduced light intensity. In a microfluidic channel, cells in the bulk flow move rapidly. At these long exposures, they appear as a faint, streaky background or "noise." Conversely, cells that are adhered or "rolling" slowly along the surface remain relatively sharp and discernible.
Automated AI Pipeline
To handle the inherent noise of MBM images, the researchers developed a two-phase automated analysis system:
- Segmentation: A modified U-Net convolutional neural network distinguishes "adhered" pixels from the blurry background.
- Classification and Tracking: The system groups pixels into individual objects and classifies them (e.g., sickle cells vs. debris) based on morphology. A custom algorithm then tracks these objects across frames to calculate adhesion duration ("survival time") and velocity.
Results and Key Insights
The study utilized diverse datasets, including sickle red blood cells (sRBCs) and CAR-T cells, to validate the pipeline. The automated system achieved an impressive R² value of 0.99 compared to manual human counting.
The Impact of Blood Dilution
A pivotal finding of this research is that standard blood dilution can mask critical disease signatures. As shown in the figure below, the significant difference in adhesion between healthy (HbAA) and sickle (HbSS) blood cells is clearly visible in whole blood but disappears entirely when the sample is diluted 1:4.
Biophysical Modeling
By tracking thousands of individual cells, the authors discovered that cell adhesion durations follow a power-law distribution. This provides a rich dataset for biophysical modeling, offering insights into how molecular bonds form and break under the shear stress of flow.
Discussion and Future Directions
While MBM is a powerful tool, it faces limitations in extremely high-density scenarios where overlapping cell trajectories can complicate tracking. Furthermore, the current implementation provides a 2D view of the surface.
Future work aims to integrate endothelialized channels—lining microfluidic paths with living vessel cells—to further enhance physiological relevance. Expanding the technique to capture 3D interactions within the flow remains a key engineering objective. Ultimately, MBM provides a low-cost, high-throughput alternative for studying cell interactions in cancer, thrombosis, and inflammatory diseases.
Reference: Goreke, U., et al. "Motion blur microscopy: in vitro imaging of cell adhesion dynamics in whole blood flow." Nature Communications (2024) 15:7058.
Summary updated