On episode 31, we discuss “Scientific Grand Challenges”– a new, rigorous approach
to evaluation of algorithms and methods. A Grand Challenge consists of a set of
data used as input to a problem of interest, and participants apply their
algorithm to that data in a blinded manner. Participant results are evaluated
against a known ground truth, and performance is quantified. This provides a
more genuine comparison of algorithms and methods than is usually possible,
leads to standard datasets for future analysis to the benefit of the field, and
often results in novel combinations of approaches that yield better results than
Our guests on this show include some of the pioneers of Grand Challenges in the
medical imaging arena.
On today’s show our crew of hosts includes:
- Stephen Alyward (special guest)
- Matt McCormick (moderator)
- Patrick Reynolds (special guest)
- Bram van Ginneken (special guest)
In 2006, Stephen Aylward left his job as a tenured associate professor of
radiology to form Kitware’s office in North Carolina. He saw open science as
they key to accelerating the pace of research in medical image analysis. He
continues to actively promote open science in his research and collaborations.
For more information, visit http://www.aylward.org/.
Bram van Ginneken is a professor in medical image analysis and chair of the
Diagnostic Image Analysis Group of Radboud University Nijmegen Medical Centre in
The Netherlands. He organized the first Grand Challenge in Medical Image
Analysis in 2007 and has been involved in 10 challenges to date. He maintains
the website http://www.grand-challenge.org/.
Patrick Reynolds is a Technical Leader at Kitware Inc with a background in
Medical Imaging and Large Data Management. He currently oversees the development
of Midas, a platform for building data sharing systems. Midas has been used to
help facilitate several Grand Challenges.
- Bram’s Grand-Challenge.org site.
- 2012 ISBI Grand Challenges.
- Michael Fitzpatrick RIRE Challenge
- Midas Challenge