Jens Rittscher – Perils of Developing Quantitative Methods for Biomedical Imaging Applications

  • Ashwani Sharma, Microsoft

By providing data sets like the PASCAL Object Recognition Database Collection and other similar data sets the computer vision community has provided an accepted benchmark that clearly defined a challenge problem for the research community at large. Although the lack of similar data sets for biomedical applications has been broadly acknowledged there are some inherent challenges that need to be addressed. Ground truth labeling of some of these very complex data sets appears to be extremely difficult. Often experts disagree and systematic approaches of finding a consensus interpretation need to be applied. The underlying variation of biological specimen is another factor that needs to be taken into account.

In this context, I will introduce the concept on edit based visualization and illustrate how it has been applied to larger scale data sets. The goal of the talk is to stimulate an exchange of ideas on this topic and discuss what lessons that we learnt from the computer vision challenge problems should be taken into account when designing a reference data set for biomedical imaging applications.

Watch Next