Applications of deep-learning based Histopathological Analysis to predict Cancer Biomarkers

Photo by Jeremy Thomas on Unsplash

Table of content

Overview

Nucleus Segmentation

Figure from [1], example of nucleus

Datasets

More information in[1]

Data description

Evaluation and Metrics

Hand-crafted techniques

UNet[3]

Illustration of U-Net semantic segmentation model[3]
Definition of the weight map
Modified training objective

Semantic segmentation for Instance segmentation(+boundary detection)

Ensembling Mask R-CNN and U-Net[4]

Results measured with IoU threshold = 0.7(oseg: over-segmented, useg: under-segmented)

Point-based Segmentation

Cancer classification

Breast Cancer Classification

Histopathologic types of breast cancer : https://en.wikipedia.org/wiki/Breast_cancer

Grade classification(Proliferation score)

http://surgpathcriteria.stanford.edu/breast/infductcabr/grading.html

Cancer Grading: Mitosis Detection

Figure in [9]

TODO: Mitosis Detection Dataset

TUmor Proliferation Assessment Challenge[11]

Demonstration of “high cellularity”, [14]

Overview of submitted approaches

TODO: DeepMitosis[9], deep learning for mitosis detection…

Cancer Grading: Tubule formation

Cancer Grading: Nuclear pleomorphism

Cancer Staging

N: CAMELYON 17 Challenge[17]

Example of approaches to predict metastasis, widely adopted in the challenge

Data

Submissions

Immunotherapy(biological therapy)

Immune phenotypes

Cancer immunity cycle

Checkpoint inhibitor, PD-1/PD-L1

Predicting PD-1/PD-L1 inhibitor prognosis by PD-L1 expression

TIL, T-cell adoptive transfer(TIL Therapy)

By Simon Caulton — Own work, CC BY-SA 3.0, https://commons.wikimedia.org/w/index.php?curid=29559885

References

16-year-old Korean, loves reading and writing about AI, DL💘. Passionate️ 🔥 about learning new technology. Contact me via LinkedIn: https://bit.ly/2VTkth7