Researchers at the Northeastern University in China have developed a deep convolutional neural network that can identify and classify different shaped red blood cells. The technology may provide cheap, easy to use devices for monitoring of patients with sickle cell disease.
Although it’s commonly assumed that sickle cell disease leads to the production of only sickle-shaped cells, in reality other shapes are also common. The types of shapes and their variety can be a biomarker of the disease, indicating how bad it’s progressing. The problem is that identifying these shapes under a microscope and grouping them into different types is too slow a process for manual work.
The algorithmic approach relies on having been provided previous images of red blood cells to learn on. It was then built so that it focuses only on red blood cells in an image, and one at a time, at which point it uses its knowledge to classify the shape of the cell.
The team verified its system against about 7,000 images of blood samples from eight sickle cell disease patients. They confirmed that the software is able to identify both oxygenated and deoxygenated red blood cells quickly and correctly.
Study in PLOS Computational Biology: A deep convolutional neural network for classification of red blood cells in sickle cell anemia…
Via: PLOS…