Workshop: Creating and Disseminating Large, Open Access, Annotated Biological Image Collections

Thursday, April 12, 18:15 - 19:15, Salon H

Organized by Bob Murphy, Carnegie Mellon University

Microscopy, image processing, and image informatics are emerging as critical core technologies in NIH's ongoing research programs in cellular, developmental and systems biology. Technology development in this area has been designated high program priority by the National Institute of General Medical Sciences (NIGMS) since 2003 ( see http://www.nigms.nih.gov/About/Council/Minutes/May15-16_2003.htm) and has also received prominent attention in several branches of the NIH Roadmap (http://nihroadmap.nih.gov). NIGMS is seeking suggestions from the community on cost-effective ways to encourage investigators in other imaging fields to work on problems in the microscopy of cells and tissues. NIGMS' interests include both the adaptation of methods already in use in other imaging fields to microscopy, and the development of new methods. They are especially interested in facilitating exploratory work by investigators in other imaging fields who wish to try their hands at microscopy.

The development of automated approaches to interpreting biomedical data is a growing and critical research area. Machine learning methods have the potential to enable understanding of the components and interactions in biological systems that are too numerous and complex for unaided human interpretation. While use of machine learning methods in biomedical research is widespread in some areas (such as genome sequence analysis), it is far less common in others. Despite significant progress over the past decade, one area where automation has not reached its full potential is in the analysis of biological data in the form of images. This is at least in part due to the relative paucity of publicly accessible biological image collections containing sufficient numbers of adequately annotated examples to enable training of machine vision systems. Widespread availability of such collections would therefore dramatically accelerate progress in the field. A possible solution is to develop mechanisms to enable investigators to create annotated image collections during the course of funded projects that make use of imaging but that currently collect only small numbers of images and rely primarily on visual interpretation. The goal of this workshop is to develop guidelines for such efforts, with the expectation that their utility would extend beyond potential funding opportunities.

The workshop will open with short presentations by panelists on relevant technologies and past experiences. These will be used to frame the subsequent discussion, which will occupy the major portion of the available time. Panelists have been chosen to cover a range of relevant expertise, with the knowledge that many other scientists may have similar experience. All scientists interested in advancing the public availability of biological image collections or in becoming involved in analyzing such collections are encouraged to attend.