Skip to main content
Advertisement
  • Loading metrics

PLoS Computational Biology Issue Image | Vol. 14(6) June 2018

  • Article
  • Metrics
  • Comments
  • Media Coverage

Classification of red blood cell shapes in flow using outlier tolerant machine learning

We investigate the shape transitions of red blood cells flowing in micro-capillaries. A convolutional neural network (CNN) transforms bright-field images of cells into a set of abstract representations. The illustration shows different shape features of the same cell highlighted by the color scheme and dependent on the applied convolution kernel. Our approach allows for a quantitative classification of each recorded cell. According to the prevailing flow conditions, the CNN reveals the formation of two stable shapes, "slippers" and "croissants". Finally, a phase diagram assessed within a range of physiolocial relevant shear rates expresses the shape transition in between these states. Kihm et al.

Image Credit: Stephan Quint

thumbnail
Classification of red blood cell shapes in flow using outlier tolerant machine learning

We investigate the shape transitions of red blood cells flowing in micro-capillaries. A convolutional neural network (CNN) transforms bright-field images of cells into a set of abstract representations. The illustration shows different shape features of the same cell highlighted by the color scheme and dependent on the applied convolution kernel. Our approach allows for a quantitative classification of each recorded cell. According to the prevailing flow conditions, the CNN reveals the formation of two stable shapes, "slippers" and "croissants". Finally, a phase diagram assessed within a range of physiolocial relevant shear rates expresses the shape transition in between these states. Kihm et al.

Image Credit: Stephan Quint

https://doi.org/10.1371/image.pcbi.v14.i06.g001