Workshop Discussion Report - Image Databases

The special character of image database systems was discussed. Traditional databases can be used to store only information derived from images while image databases provide image retrieval as well. Some image database systems also provide image reprocessing after retrieval.

The current state of image database systems for biological research was reviewed. The SciMagix SIMS system is the primary commercial image database for biomedical purposes that provides both image retrieval and image reprocessing. The Cellomics database focuses on storing information derived from images (the end-user cannot reprocess images). A number of companies have in-house image databases and Lawrence Berkeley Laboratories is working on an image database using ObjectStore.

The roles of people involved with image databases were delineated

• database administrator
  • deals with resource issues, backups
• software component developer
  • provides image processing/analysis algorithms (local or remote)
  • does not require access to images
  • system provides mechanism for retrieval of images and associated fields
  • system provide ability to return images and associated fields
• application developer
  • combines software components to create applications
  • does not require access to images
• end user
  • enters images (implicitly via image acquisition system or explicitly via loading files)
  • has access to images
  • requires ability to create temporary databases
  • derives value from images entered and database queries

1. An implementation-independent standard for exchange of images is needed, either between image databases or between an image database and an analysis client. For this purpose, there is a need to define a minimal, required set of fields describing image records (dimensions, bit depth).
2. A standard database schema for storing image annotation is needed. The workshop recommends that draft microscope image description tables such as those presented at the workshop by the Murphy group be combined with additional comments from other laboratories. LIMS (Laboratory Information Management System) should be consulted for additional experiment description fields. The standard should recognize the need to create locally-defined tables to augment the standard schema and should provide a standard mechanism for exporting descriptions of locally-defined fields.
3. The ISAC Data Standards Committee can play a major role in these efforts.