@article{fdi:010087700,
  title = {{C}lassification of red cell dynamics with convolutional and recurrent neural networks : a sickle cell disease case study},
  author = {{D}arrin, {M}. and {S}amudre, {A}. and {S}ahun, {M}. and {A}twell, {S}. and {B}adens, {C}. and {C}harrier, {A}. and {H}elfer, {E}. and {V}iallat, {A}. and {C}ohen-{A}ddad, {V}. and {G}iffard-{R}oisin, {S}ophie},
  editor = {},
  language = {{ENG}},
  abstract = {{T}he fraction of red blood cells adopting a specific motion under low shear flow is a promising inexpensive marker for monitoring the clinical status of patients with sickle cell disease. {I}ts high-throughput measurement relies on the video analysis of thousands of cell motions for each blood sample to eliminate a large majority of unreliable samples (out of focus or overlapping cells) and discriminate between tank-treading and flipping motion, characterizing highly and poorly deformable cells respectively. {M}oreover, these videos are of different durations (from 6 to more than 100 frames). {W}e present a two-stage end-to-end machine learning pipeline able to automatically classify cell motions in videos with a high class imbalance. {B}y extending, comparing, and combining two state-of-the-art methods, a convolutional neural network ({CNN}) model and a recurrent {CNN}, we are able to automatically discard 97% of the unreliable cell sequences (first stage) and classify highly and poorly deformable red cell sequences with 97% accuracy and an {F}1-score of 0.94 (second stage). {D}ataset and codes are publicly released for the community.},
  keywords = {},
  booktitle = {},
  journal = {{S}cientific {R}eports - {N}ature},
  volume = {13},
  numero = {1},
  pages = {[12 ]},
  ISSN = {2045-2322},
  year = {2023},
  DOI = {10.1038/s41598-023-27718-w},
  URL = {https://www.documentation.ird.fr/hor/fdi:010087700},
}