RBC Differential
Computer Vision Analysis of Red Blood Cell Morphology
We developed the RBC-diff, a computer vision system for identifying red cell abnormalities. Select Select a patient case below, to view their blood smear, and identify dysmorphologous red cells! Based on our Blood Advances 2023 paper.
Sample Blood Smears
Select a blood smear image to view its RBC morphology distribution:
About This Method
What it does: The RBC-diff uses computer vision to detect abnormal cells. Individual cells are first detected using a simple binary filter. The model then calculates geometric features for each cell, and passes it to a machine learning model for classification. This model has been robustly validated at the single-cell and smear-level against expert hematopathologist labels.
RBC Shape Categories:
- Microcytes: Small cells, often seen in iron deficiency and thalassemia
- Macrocytes: Large cells, often seen in B12/folate deficiency
- Elliptocytes: Elliptic cells, often due to genetic conditions such as hereditary elliptocytosis
- Schistocytes: Fragmented cells, indicative of hemolytic stress such as thrombotic microangiopathies
- Sickle Cells: Crescent-shaped cells, due to sickle cell disease
- Spiculated Cells: Cells with spiked edges (echinocytes and acanthocytes), due to many conditions, such as liver disease
- Teardrops: Teardrop-shaped cells, seen in myelofibrosis and other conditions
Clinical Applications: Automated RBC quantitation allows for more rapid and objective diagnosis, prognosis, and monitoring of hematologic conditions.
Code Availability: The code and training set images for the RBC morphology classification model are available from the Blood Advances 2023 publication, as supplementary materials.
Learn more: Our Research | Blood Advances 2023 Paper | Publications