We are happy to announce that our work in enriching lung cancer collections with AI-derived annotations has been published in Nature Scientific Data!
Krishnaswamy, D., Bontempi, D., Thiriveedhi, V.K. et al. Enrichment of lung cancer computed tomography collections with AI-derived annotations. Sci Data 11, 25 (2024). Enrichment of lung cancer computed tomography collections with AI-derived annotations | Scientific Data
This paper introduces how we used cloud-enabled workflows for generating annotations for two lung cancer CT collections in IDC. We annoated a subset of the National Lung Screening Trial (NLST) collection and the NSCLC-Radiomics collection. We used pre-trained models to annotate thoracic organs-at-risk and create slice level annotations. Additionally, we also extracted shape features of these regions to interpret the annotations. Check out the ingested data in IDC here and an example here!
Though authors are now starting to make their code and data reproducible, many times information about how to interact with the data is missing. Therefore, we provide multiple demonstrations of how to download, interact and visualize the annotations in this GitHub repository.
We also developed a Looker Studio Dashboard for interactive analysis, where you can filter for shape features, look at histogram of values, and click on the associated OHIF urls to examine patients:
Lastly, we demonstrate how to interact with, download and visualize the annotations we generated in a Colab notebook. Here’s an example of using the bokeh package to visualize the sphericity feature – click on a point to open up the associated series in the OHIF viewer!
