Mapping cell-to-tissue graphs across human placenta histology whole slide images using deep learning with HAPPY

Vanea, Claudia; Džigurski, Jelisaveta; Rukins, Valentina; Dodi, Omri; Siigur, Siim; Salumäe, Liis; Meir, Karen; Parks, W. Tony; Hochner-Celnikier, Drorith; Fraser, Abigail; Hochner, Hagit; Laisk, Triin; Ernst, Linda M.; Lindgren, Cecilia M.; Nellåker, Christoffer

doi:10.6082/5zdpm-cac60

Published March 28, 2024 | Version v1

Journal article Open

Mapping cell-to-tissue graphs across human placenta histology whole slide images using deep learning with HAPPY

1. University of Oxford
2. University of Tartu
3. Hadassah Hebrew University Medical Center
4. Tartu University Hospital
5. University of Toronto
6. University of Bristol
7. Hebrew University of Jerusalem
8. University of Chicago

Accurate placenta pathology assessment is essential for managing maternal and newborn health, but the placenta's heterogeneity and temporal variability pose challenges for histology analysis. To address this issue, we developed the 'Histology Analysis Pipeline.PY' (HAPPY), a deep learning hierarchical method for quantifying the variability of cells and micro-anatomical tissue structures across placenta histology whole slide images. HAPPY differs from patch-based features or segmentation approaches by following an interpretable biological hierarchy, representing cells and cellular communities within tissues at a single-cell resolution across whole slide images. We present a set of quantitative metrics from healthy term placentas as a baseline for future assessments of placenta health and we show how these metrics deviate in placentas with clinically significant placental infarction. HAPPY's cell and tissue predictions closely replicate those from independent clinical experts and placental biology literature.

Data availability

The datasets generated for training and validating each deep learning model along with trained model weights are available for download at the Google Drive link: https://tinyurl.com/happyplacenta or from Zenodo: 10.5281/zenodo.10535021 with no restrictions. Instructions can be found in the GitHub readme at: https://github.com/Nellaker-group/happy. The two histology slides used for graph model training are available for download under CC BY 4.0 from the BioImage Archive at 10.6019/S-BIAD1045. The remaining in-house placenta histology slides and clinical data are not made available in accordance with existing research ethics committee approvals and data transfer agreements. Pretrained ImageNet weights for the RetinaNet and ResNet-50 models were downloaded in code via PyTorch from https://download.pytorch.org/models/resnet101-5d3b4d8f.pth and https://download.pytorch.org/models/resnet50-0676ba61.pth and will be downloaded programmatically on first model use. Source data are provided with this paper.

Code is available at the following GitHub repository https://github.com/Nellaker-group/happy and at https://doi.org/10.5281/zenodo.10529239.

Files