BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Numerical Methods for CT Reconstruction with Unknown Geometry Para
 meters - James Nagy (Emory University)
DTSTART:20230327T130000Z
DTEND:20230327T135000Z
UID:TALK198202@talks.cam.ac.uk
DESCRIPTION:Computed tomography (CT) techniques are well known for their a
 bility to produce high quality images needed for medical diagnostic purpos
 es. Unfortunately standard CT machines are extremely large\, heavy\, requi
 re careful and regular calibration\, and are expensive\, which can limit t
 heir availability in point-of-care situations. An alternative approach is 
 to use portable machines\, but parameters related to the geometry of these
  devices (e.g.\, distance between source and detector\, orientation of sou
 rce to detector) cannot always be precisely calibrated\, and these paramet
 ers may change slightly when the machine is adjusted during the image acqu
 isition process. In this work\, we describe the nonlinear inverse problem 
 that models this situation\, and discuss algorithms that can jointly estim
 ate the geometry parameters and compute a reconstructed image. In particul
 ar\, we propose a hybrid machine learning and block coordinate descent (ML
 -BCD) approach that uses an ML model to calibrate geometry parameters\, an
 d uses BCD to refine the predicted parameters and reconstruct the imaged o
 bject simultaneously. Numerical experiments illustrate that our new method
  can efficiently improve the accuracy of both the image and geometry param
 eters. This is joint work with Chang Meng.
LOCATION:Seminar Room 1\, Newton Institute
END:VEVENT
END:VCALENDAR