Researchers from Department of Signal Processing win an ultrasound segmentation challenge
Alessandro Foi , Matteo Maggioni, Antonietta Pepe, and Jussi Tohka from the Department of Signal Processing have won the “Biometric measurements from fetal ultrasound images” challenge organized in conjunction of prestigious ISBI (IEEE International Symposium on Biomedical Imaging) 2012 conference.
The task in this image segmentation challenge was to automatically measure standard obstetric biometric parameters from 2D fetal ultrasound images of varying image quality taken on fetuses of different gestational ages. The automatic measurements were compared to manual measurements based on ultrasound images made by medical experts.
Several international teams participated to the challenge, including teams representing University of Oxford (United Kingdom), Russian Academy of Sciences, Ecole Polytechnique Fédérale de Lausanne (Switzerland), City University of Hong Kong, Technical University of Cluj-Napoca (Romania), and Commonwealth Scientific and Industrial Research Organisation (Australia).
The winning method by TUT signal processing researchers “Head contour extraction from fetal ultrasound images by difference of Gaussians revolved along elliptical paths” was a novel cost-function based segmentation approach. In other words, the optimal segmentation, to which the biometric measurements can be based on, was found by minimizing a cost function. The minimization problem in itself is challenging. It was solved by using a stochastic multiscale multistart method.
Link to the challenge web-site : http://www.biomedicalimaging.org/index.php/programme/isbi-challenges/13-challenges/53-contest-workshop-challenge-us
More information:
Academy research fellow Alessandro Foi (alessandro.foi@tut.fi)
Academy research fellow Jussi Tohka (jussi.tohka@tut.fi)