Skip to Content
COVID-19 Resources
Cancer Imaging Program (CIP)
Contact CIP
Show menu
Search this site
Last Updated: 10/28/16

Breakout Group 2 — Summary and Discussion

Linkage of image databases with other forms of biological knowledge

Curtis Langlotz (Leader)
Allan Johnson (Co-leader)
Zhan Zhang
John Fryman
Yanxi Liu
Daniel Barbara
Kathleen Robinette
Joseph Wang
Fei Xu
Ann Menkens
Hanan Samet
Larry Kerschberg
Amarnath Gupta
Beverly Meadows
Rich Cumberlin
Robert Tinker
Yannick Kergosian
John Haller
Sanjay Ramka
Ed Staab
Edward Herskovits

Discussion Framework:

Software Design Metaphor

Definitions and context
3 "Use cases"
What linkages are required
Gap/requirements analysis

Types of Imaging Data

Clinical images (humans)
Small animal images
Histology and pathology
3D chemical structure
Other microscopy (confocal, EM, etc.)
In situ hybridization

Non-Imaging Biological Data

Thesauri, lexicons, vocabulary, ontology
Medical literature, entrez databases
Physiologic models and biochemical pathways
Electronic medical records
Drug databases
Protein classification and function
Video surgery

Normal human and visible human

Use Case A: From a Clinical Radiographic Image

What is the diagnosis given imaging features?
What are the radiographic features to look for, given a specific diagnosis?
What are the most salient items from the electronic medical record?
What is optimal patient management/treatment?
What is the molecular biology of the disease?
What is the relevant medical literature?
What are relevant areas for further research?

Use Case B: From an image of a tumor model in a small animal

What is the connection to the human model?
What is the connection to gene expression?
What correlative information is available from other images (histology, EM, histochemistry)?
What is the gene expression and distribution in humans?

Use Case C: Based on a population of images of a specific disease

More information on the disease or phenotype?
Similar images from other databases/tumors?
What human populations (as identified by geography, genetic markers, body shape, etc) are susceptible?

Use Case Requirements

Feature extraction: pixels features, concept (eg, putamen, mass effect)
Common ontology/thesaurus of data elements, features, and concepts
Problem solving workflow
Automatic retrieval of similar images based on texture, shape, position.

The BIRN Model

Clinical Images connect to Concept Assignment connects to Other Biological Knowledge which may be in a wrapper; and Concept Assignment connects to UMLS

Database Linkage Model

Clinical images connect to Database semantics connect to Ontology/thesaurus while Other biological knowledge connects to another Database semantics and thus to the Ontology/thesaurus

Barriers to Linkage (1)

No common and complete imaging data elements
None of the databases talk the same language: different identifiers for the same concept
None of the radiologists talk the same language: different meanings for the same concept
People and computers talk different languages: (eg, "mass effect")

Barriers to Linkage (2)

Semantics (and APIs) of many existing databases are unpublished — reverse engineering required
No thesaurus/ontology for non-clinical data (genes, proteins), internal creation and maintenance of a thesaurus required


Support efforts development and curation of imaging common data elements
Support efforts to enhance richness of CDEs with description logic and representative images
Support harmonized thesaurus/ontology of non-imaging data elements (UMLS for gene and protein data)
Support development of tools to retrieve by image content