The program will be available shortly. Please check back later.
Presenter: Professor Lars Kai Hansen, Head of Section Cognitive Systems, Department of Applied Mathematics and Computer Science - DTU Compute.
Abstract: Learning from labeled and unlabeled data is key to many advances in machine learning. Self-supervised learning has made strides in natural language processing, speech and video processing and is emerging as a powerful approach to bio-medical applications. Theoretical models have been analysed since the early days of machine learning. I will review important theoretical insights,present some new results and point to open problems.
Presenter: Associate Professor Irina Voiculescu, Department of Computer Science, Oxford University.
Abstract: Screening for medical conditions has been the conventional reason for segmenting anatomical features as either contours or binary masks. Clinical tasks often involve the need for further measurements to be taken from such segmented features.
We explore the clinical and scientific impact of taking specific measurements of the anatomy without segmenting it at all, simply from identifying key point landmarks in each image. Moreover, we discuss how to obtain an uncertainty measurement for each landmark, and its effect on improving clinicians' trust in the landmark predictions. We illustrate the method on a few concrete examples of significant landmarks in xrays, using both clinical examples and public datasets.
In compliance with GDPRconsent requirements, presentations given in a Visual Intelligence context maybe recorded with the consent of the speaker. All recordings are edited toremove all faces, names and voices of other participants. Questions and commentsby the audience will hence be removed and will not appear in the recording. With the freely given consent from the speaker, recorded presentation may beposted on the Visual Intelligence YouTube channel.