Abstract #2057

Generalisability of Image Quality Transfer: Can we approximate in-vivo human brains from dead monkey brains?

Aurobrata Ghosh¹, Viktor Wottschel², Enrico Kaden¹, Jiaying Zhang¹, Hui Zhang¹, Stamatios N. Sotiropoulos³, Darko Zikic⁴, Tim B. Dyrby⁵, Antonio Criminisi⁴, and Daniel C Alexander¹

¹Centre for Medical Image Computing, University College London, London, United Kingdom, ²Institute of Neurology, University College London, London, United Kingdom, ³Oxford Centre for Functional Magnetic Resonance Imaging of the Brain (FMRIB), University of Oxford, Oxford, United Kingdom, ⁴Microsoft Research, Cambridge, United Kingdom, ⁵Danish Research Centre for Magnetic Resonance, Centre for Functional and Diagnostic Imaging and Research, Copenhagen University Hospital Hvidovre, Hvidovre, Denmark

The Image-Quality Transfer (IQT) framework enhances low quality images by transferring information from high quality images acquired on expensive bespoke scanners. Although IQT has major potential in medical imaging, one key question is its dependence on training data. We demonstrate the generalisability of IQT used for super-resolution by showing that reconstruction of in-vivo human images degrades minimally from training on human data from the same study, to data from a different demographic and imaging protocol, to data from fixed monkey brains. Remarkably, a patchwork of fixed monkey brain image-pieces is hardly distinguishable from a reconstruction using pieces of human data.

This abstract and the presentation materials are available to members only; a login is required.

Join Here