Oral Cell Screening Project

Code of paper A Deep Learning based Pipeline for Efficient Oral Cancer Screening on Whole Slide Images

Pre-print version on arXiv

Usage instruction

Each sub-directory contains a separate README instruction. File paths in the code might need to be changed before running.

OralCellDataPreparation/ – Nucleus Detection (ND) and Focus Selection (FS) modules trained on our data. This will prepare all nucleus patches for classification. It can be tuned to better performance on a new dataset by the code in two directories below:
- NucleusDetection/ – to customise the ND module.
  - VOC2007/ - to prepare training data for the ND module.
- FocusSelection/ – to customise the FS module.
Classification/ – Classification module.

Example results

Left: pineline on pap-smear data; Right: pineline on liquid-based data

Dependencies

oralscreen_env.yml includes the full list of packages used to run the experiments. Some packages might be unnecessary.

Citation

Please cite our paper if you find the code useful for your research.

J. Lu et al., “A Deep Learning based Pipeline for Efficient Oral Cancer Screening on Whole Slide Images,” International Conference on Image Analysis and Recognition, 2020, LNCS, vol 12132.

@inproceedings{OralScreen,
  title = {A {{Deep Learning Based Pipeline}} for {{Efficient Oral Cancer Screening}} on {{Whole Slide Images}}},
  booktitle = {Image {{Analysis}} and {{Recognition}}},
  author = {Lu, Jiahao and Sladoje, Nataša and Runow Stark, Christina and Darai Ramqvist, Eva and Hirsch, Jan-Michaél and Lindblad, Joakim},
  date = {2020},
  pages = {249--261},
  publisher = {{Springer International Publishing}},
  location = {{Cham}},
  doi = {10.1007/978-3-030-50516-5_22},
  isbn = {978-3-030-50516-5},
  langid = {english},
  series = {Lecture {{Notes}} in {{Computer Science}}}
}

Acknowledgement