We provide fully trained UNet segmentation weights for WSI IHC synovial tissue which can be used as the first step in an automated image analysis pipeline. It is robust to common WSIs artefacts, clinical centre/scanner batch effect and can be used on different types of IHC stains. It can be used as is, or fine-tuned on any IHC musculoskeletal dataset, removing the need for manual tissue segmentation by pathologists.
A total of 164 patients, fulfilling the 2010 American College of Rheumatology/European Alliance of Associations for Rheumatology (EULAR) classification criteria for RA were recruited to the R4RA clinical trial from 20 European centers [15] [7]. Patients underwent ultrasound-guided synovial biopsy of a clinically active joint. Samples were then fixed in formalin, embedded in paraffin, cut with microtome and stained with the relevant IHC stains: IHC CD20 (B cells), IHC CD68 (macrophages) and IHC CD138 (plasma cells) [7]. Samples were then placed on glass slides and scanned into Whole Slide Image (.ndpi format) with digital scanners under 40x or 20x objectives. Below we show representative examples of the three IHC stains used to train the UNet algorithm:
Below we show some segmentation results. The algorithm is robust to many WSIs artefacts and to the fragmeted nature of synovial tissue samples and the low contrast inherent to IHC dyes:
If this code or weights are useful to you, please consider citing:
@inproceedings{gallagher2023automated,
title={Automated segmentation of rheumatoid arthritis immunohistochemistry stained synovial tissue},
author={Gallagher-Syed, A and Khan, A and Rivellese, F and Pitzalis, C and Lewis, MJ and Slabaugh, G},
booktitle={27th Conference on Medical Image Understanding and Analysis 2023},
pages={76}
}
Gallagher-Syed A., Khan A., Rivellese F, Pitzalis C., Lewis M. J., Slabaugh G., Barnes M., "Automated segmentation of rheumatoid arthritis immunohistochemistry stained synovial tissue", Medical Image Understanding and Analysis, Aberdeen. 2023. Conference abstract paper.