Three-dimensional morphological characterization of blood droplets during the dynamic coagulation process

Area of Science:

• Biomedical Engineering
• Medical Imaging
• Artificial Intelligence in Medicine

Background:

• Blood coagulation is a complex process vital for hemostasis.
• Accurate, quantitative monitoring of coagulation is crucial for clinical diagnostics.
• Current methods may lack the precision and dynamic range needed for real-time analysis.

Purpose of the Study:

• To develop and validate an AI-driven method for quantitative characterization of whole blood coagulation.
• To integrate optical coherence tomography (OCT) with deep learning frameworks (U-Net, VGG-Net).
• To assess the accuracy and segmentation capabilities of the proposed approach.

Main Methods:

• Utilized a convolutional neural network integrating OCT imaging with U-Net and VGG-Net architectures.
• Employed VGG-Net for identifying blood droplets across distinct coagulation stages (drop, gelation, coagulation).
• Applied U-Net for segmenting uncoagulated, coagulated blood portions, and background.

Main Results:

• VGG-Net achieved up to 99% accuracy in identifying blood droplets during coagulation.
• U-Net effectively segmented different blood states and background, enabling precise volume measurements.
• Quantitative parameters like uncoagulated and coagulated segment volumes were successfully derived.

Conclusions:

• The integrated OCT and deep learning approach provides accurate, quantitative characterization of blood coagulation.
• This method demonstrates significant potential for future clinical diagnostics and real-time hemostasis analysis.
• AI-powered image analysis offers a promising tool for understanding dynamic biological processes like coagulation.