BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Memento EPFL//
BEGIN:VEVENT
SUMMARY:Adaptive signal representation for accelerated dynamic MRI
DTSTART:20130115T140000
DTSTAMP:20260408T082719Z
UID:11f66dc3bfaf4d373ff82287b715c10d4eacba0a5d186812e924f13b
CATEGORIES:Conferences - Seminars
DESCRIPTION:Dr. Mathews Jacob\, University of Iowa\nCardiac MRI schemes su
 ch as myocardial perfusion and viability imaging play key roles in evaluat
 ing both ischemic and non-ischemic heart disease. The need to reliably det
 ect subtle lesions requires good in-plane spatial resolution\, good spatia
 l coverage\, high temporal resolution\, long breath-hold duration and good
  contrast to noise ratio. Since these contradictory goals are often diffic
 ult to realize\, clinicians are often forced to compromise on spatio-tempo
 ral resolution and coverage\nThe main focus of this talk is to introduce a
 daptive signal representations to considerably improve the state of the ar
 t in dynamic MRI. I will start by introducing blind linear models\, where 
 the basis functions and the coefficients of the representation are estimat
 ed from highly under-sampled MRI data. The recovery is posed as an optimiz
 ation problem\, which is solved using a fast augmented Lagrangian algorith
 m. This framework is then extended to blind compressive sensing\, where th
 e dictionary basis functions and the sparse coefficients are estimated fro
 m heavily under-sampled data. I will also introduce a novel algorithm for 
 dictionary learning from under-sampled data. Unlike classical algorithms\,
  the proposed scheme can handle arbitrary dictionary constraints\; I will 
 also illustrate the utility in using alternate dictionary constraints to f
 urther improve the recovery.
LOCATION:AAC114
STATUS:CONFIRMED
END:VEVENT
END:VCALENDAR