Related Experiment Video
Updated: Jun 30, 2026
08:04
Analyzing Platelet Subpopulations by Multi-color Flow Cytometry
Published on: June 10, 2025
Associating the phenotypic expression of platelets with disease type through image-based single-cell profiling
Huidong Wang1, Masako Nishikawa2, Yuqi Zhou1
1Department of Chemistry, The University of Tokyo, Tokyo, 113-0033, Japan.
Thrombosis Research
|June 28, 2026
Summary
Researchers developed AI models to classify platelet phenotypes from images, identifying disease-associated signatures. These models can predict thrombotic progression days in advance, aiding early disease detection and risk assessment.
Area of Science:
- Hematology
- Biomedical Imaging
- Artificial Intelligence
Background:
- Platelets are crucial for hemostasis and thrombosis, implicated in various diseases like cardiovascular disorders, infections, and cancer.
- Disease-associated platelet phenotypes are subtle, transient, and difficult to resolve at the single-cell level, hindering comprehensive study.
- Current understanding of platelet dysfunction in disease lacks detailed phenotypic characterization.
Purpose of the Study:
- To develop and validate artificial intelligence (AI) models for classifying platelet phenotypes from high-resolution images.
- To identify disease-associated platelet signatures using advanced imaging and deep learning techniques.
- To assess the potential of AI-driven platelet analysis for early disease detection and risk stratification.
Main Methods:
- Collected whole-blood samples from patients and healthy volunteers.
- Utilized optofluidic imaging with an optical frequency-division-multiplexed (FDM) microscope and microfluidic chip for high-resolution platelet imaging.
- Trained convolutional neural network (CNN) models to classify platelet phenotypes across disease categories and performed feature-importance analysis.
Main Results:
- Developed three CNN models achieving up to 81.3% accuracy in classifying platelet images by disease category.
- Demonstrated that CNN models could predict thrombotic progression up to 7 days prior to clinical detection through longitudinal analysis.
- Identified texture-related descriptors as key features driving CNN-based platelet characterization, accounting for 54.5% of importance.
Conclusions:
- Image-based analysis of circulating platelets can effectively capture disease-associated phenotypic signatures.
- AI-driven platelet phenotyping shows promise for complementing existing clinical workflows.
- This approach may enhance prediagnosis and enable early assessment of thrombotic risk.
