Deep Histology Classification

Research on deep learning for histology. The full pipeline is used to decompose large tif histology slides into smaller images (tiles) that can be used for training a deep convolutional neural network which are then used to generate heatmaps of pathology hot spots on input images.

Tiling

Tool that is used for preprocessing ndpi images into tiff images for entire directories and also
generates training tiles from images annotated using QuPath. See Tiling directory.

Training

Tool that is used to train deep convolutional nerual networks on a tile dataset generated from the Tiling tool. Trained models are then used as input for the Evaluation tool. See Training directory.

Evaluation

Tool that tiles a tissue image and uses a pre-trained deep learning model to classify the tiles as well as show statistics between control and experimental groups. See Evaluation directory.

Heatmap sample of model predictions

Class specific sample of model predictions

Acknowledgements

This work was supported by the John Templeton Foundation grant number 61174.