BEGIN:VCALENDAR VERSION:2.0 PRODID:-//Talks.cam//talks.cam.ac.uk// X-WR-CALNAME:Talks.cam BEGIN:VEVENT SUMMARY:Task Oriented Reconstruction using Deep Learning - Ozan Öktem (KT H - Royal Institute of Technology \; Karolinska Institute) DTSTART:20171031T095000Z DTEND:20171031T104000Z UID:TALK94108@talks.cam.ac.uk CONTACT:INI IT DESCRIPTION:Co-Author: Jonas Adler \;

Machine learning has be en used if image reconstruction for several years\, mostly driven by the r ecent advent of deep learning. Deep learning based reconstruction methods have been shown to give good reconstruction quality by learning a reconstr uction operator that maps data directly to reconstruction. These methods t ypically perform very well when performance is measured using classical qu antified\, such as the RMSE\, but they tend to produce over-smoothed image s\, reducing their usefulness in applications. \;

We propos e a framework based on statistical decision theory that allows learning a reconstruction operator that is optimal with respect to a given task\, whi ch \;can be segmentation of a tumor or classification. In this framew ork\, deep learning is used not only to solve the inverse problem\, but al so to simultaneously learn how to use the reconstructed image in order to complete an end-task. We demonstrate that the framework is computationally feasible and that it can improve human interpretability of the reconstruc tions. We also suggest new research directions in the field of data driven \, task oriented image reconstruction. \;

Related publicati ons:
http://arxiv.org/abs/1704.04058 (accepted for publicati on in Inverse Problems)
http://arxiv.org/abs/1707.06474 (submitted to IEEE Transaction for Medical Imaging) LOCATION:Seminar Room 1\, Newton Institute END:VEVENT END:VCALENDAR