Hes-so Valais

Benchmarking Medical Visual Information Systems

Benchmarks or large-scale evaluation campaigns to compare research techniques based on the same grounds have proven extremely useful for research fields that are on the edge from basic research to first products in domains such as information retrieval and automatic translation. For the management of complex media such as images, videos, free text, and combinations of these, several benchmarks have started in the past five years with varying objectives. The medical domain just like many other domains produces enormous amounts of visual data that are connected with other data in the electronic patient record.

ImageCLEFmed has started to create a benchmark for these visual retrieval techniques and the impact of the campaign could be much more important with a solid infrastructure and analysis.

BeMeVIS (Benchmarking Medical Visual Information Systems) will create such an infrastructure for ImageCLEF and other benchmarks, and will increase the impact of the benchmarks by analyzing the workflow and developing a software support for all important steps of the workflow.

Besides this, a retrieval prototype for medical visual information retrieval will be developed based on the results of the best-performing techniques. This prototype will allow addressing partners in the medical domain searching for solutions for managing visual medical information.