engrodawood/albrt

Cellular Composition Prediction in Routine Histology Images

ALBRT

Cellular Composition Prediction in Routine Histology Images

Cellular composition prediction, i.e., predicting the presence and counts of different types of cells in the tumor microenvironment from a digitized image of a Hematoxylin and Eosin (H&E) stained tissue section can be used for various tasks in computational pathology such as the analysis of cellular topology and interactions, subtype prediction, survival analysis, etc. In this work, we propose an image-based cellular composition predictor ALBRT which can accurately predict the presence and counts of different types of cells in a given image patch. ALBRT, by its contrastive-learning inspired design, learns a compact and rotation-invariant feature representation that is then used for cellular composition prediction of different cell types. It offers significant improvement over existing state-of-the-art approaches for cell classification and counting. The patch-level feature representation learned by ALBRT is transferrable for cellular composition analysis over novel datasets and can also be utilized for downstream prediction tasks in CPath as well.

Dataset

For model training and validation PanNuke dataset was used.

The pretrained weights for all three folds of PanNuke dataset can be downloaded from Here

For assessing model generalization, NuCLS dataset consisting of 1, 744 field of views (FOVs) partitioned into 5 pre-defined train and validation folds was used.

Proposed Model for cellular composition prediction

ALBRT test prediction

Comparison of ALBRT and HoVer-Net prediction with ground truth for a set of test images. Images are overlaid with annotatedcellular boundaries. Below each image true cellular counts, HoVer-Net and ALBRT predicted cellular counts for different types of cells are shown.

Saliency Map Analysis

ALBRT captures interclass and intraclass cellular variability

License

Note that the PanNuke dataset is licensed under Attribution-NonCommercial-ShareAlike 4.0 International , therefore ALBRT weights are also under the same license. The source code of ALBRT is released under MIT-CC-Non-Commercial license.

albrt's Issues

Request for ALBRT Model Training Code

Hello,

I hope this message finds you well. I am very interested in the ALBRT model and its applications. After reviewing the paper and related code, I am eager to gain a deeper understanding of the model's training process.

I am currently working on a project that involves similar tasks in the field of computer vision and histology images, and I believe that studying the ALBRT training code would greatly benefit my research.

If possible, could you please share the code related to the training process of the ALBRT model? I understand that sharing code can be subject to certain conditions or considerations, and I would be more than willing to comply with any requirements you may have.

Thank you for considering my request, and I look forward to your response. If you have any questions or need further information, please feel free to ask.