Flexibility and Search-ability: Support for Quantitative Analysis and Content-based Image Retrieval of Tissue Microarrays
Wenjin Chen ; UMDNJ;
Content:
High throughput Tissue Microarray (TMA) technology is gaining wide acceptance by investigators throughout cancer research community. Future progress in TMA-based research is reliant upon the capacity to archive, manage, and share TMA related information among research groups and institutions. We recently developed a Tissue Microarray Repository (TMR) module to organize and manage TMA related information and images. The new TMR interface facilitates the population of distributed databases with new datasets including image metrics and correlated profiles in multi-user environments. While we are integrating the existing and emerging TMA data exchange standards into the TMR we are also establishing new metadata sets to support a range of image-based feature measurements that are being developed as part of an NIH funded research project.
Technology:
The TMR framework is being developed to support seamless extensions, or plug-in applications, for performing quantitative image analysis while providing access to flexible database structure and image archive for storage and retrieval. The system is developed using JAVA and a Oracle 10g database with a denormalized structure design.
Design:
The TMR module is composed of a physical specimen layer, a digital specimen layer and a quantification layer. The quantification layer is designed to provide flexibility in accommodating heterogeneous data forms that are generated from clinical, biological and image-based TMA research. In the development stages of each project, the researchers have maximal flexibility in designing and exploring different metadata forms. As the datasets and analytical tools continue to mature, these flexible structures can be easily migrated into their permanent forms so as to provide optimal efficiency in allowing Grid-enabled sharing and searching of these data and meta-data.
Results:
We have recently migrated into TMR a content-based image retrieval application which enables one to query a gold-standard database of over 3000 previously diagnosed breast cancer TMA discs and provide decision support for scoring and comparative analysis.
Conclusion:
As the TMA related Common Data Elements are still being developed and evolving, the TMR framework provides a reliable solution for accommodating heterogeneous needs of TMA research community. Flexibility and share-ability of the data are both taken into consideration in the design of the framework.
