The W-net Deep Learning Model for Image Reconstruction

Share

Magnetic resonance imaging (MRI) is a diagnostic imaging modality that provides excellent image contrast that leads to diagnoses of several conditions, such as tumours and strokes. The major drawback of MRI exams is their long acquisition time. An MRI exam usually lasts between 30 to 60 minutes per patient. These long acquisition times, combined with the high costs (~$3,000,000) and space requirements to install a scanner and periodic maintenance, cause MRI exams to be expensive and create wait times that make it less accessible. An MRI exam's average price is over $700, and nearly two million MRI exams are done in Canada yearly, making it a billion-dollar industry.

In this talk, Dr. Souza will cover the basics of deep learning methods for image reconstruction. He will present a deep learning model called the W-net that can be used to make MRI exams up to 20 times faster. Dr. Souza will also briefly illustrate the use of the W-net to improve other imaging applications, such as JPEG image decompression and low-dose computed tomography reconstruction.

--------------------------------------------------------------------------------------------------------------------------------------------------------------------

Roberto Souza is an Assistant Professor at the Electrical and Computer Engineering Department at the University of Calgary, Canada, since July 2020. He has a B.Sc. in electrical engineering from the Federal University of Pará (2012), an M.Sc. (2014) and Ph.D. (2017) both in computer engineering from the University of Campinas (UNICAMP). Before becoming an Assistant Professor, he worked for three years as a postdoctoral scholar at the University of Calgary. He has international experience, having worked as an intern at the Grenoble Institute of Technology, France, and the University of Pennsylvania, United States. Dr. Souza has extensive expertise in image processing and machine learning. His research is currently focused on smart strategies for data integration and data mining.



  Date and Time

  Location

  Hosts

  Registration



  • Date: 24 Nov 2020
  • Time: 12:00 PM to 01:00 PM
  • All times are America/Edmonton
  • Add_To_Calendar_icon Add Event to Calendar

Virtual session information will be sent to registrants the day before the event

  • Starts 02 November 2020 04:00 PM
  • Ends 23 November 2020 12:00 PM
  • All times are America/Edmonton
  • No Admission Charge