PRESENTATIONS (2)
Student presentation:
“Data-driven respiratory motion correction of myocardial perfusion SPECT
using a convolutional neural network”
Respiratory motion leads to degradation of image quality in cardiac SPECT
myocardial perfusion studies. While respiratory gating can minimize this
effect, many implementations involve the use of additional equipment and
manual alignment of the gated images which is time consuming for the
technologist. This work aims to automatically correct cardiac SPECT data
for respiratory motion using data-driven respiratory gating combined with
motion parameter estimation using a convolutional neural network. This
approach does not require images to be reconstructed beforehand, allowing
it to be seamlessly integrated into an imaging workflow.
By Dylan Malenfant
Department of Physics, Carleton University
Supervisor: Dr. Glenn Wells, University of Ottawa Heart Institute
Member presentation:
“From visualization to image processing: a physicist's foray into
professional software development”
Radiologists, nuclear medicine physicians, and their technologists rely
heavily on core imaging tools that allow them to visualize, process and
quantity diagnostic medical images to help referring physicians make
diagnoses and guide therapy decisions for their patients. These products
are widely available either as commercial or open-source offerings which
may make it easy for one who consumes them to overlook the details of how
they actually transform raw image data into the images that are seen on a
computer display.
In this talk I'll take a look "under the hood" of UniSyn Molecular
Imaging, the application to which I contribute at Convergent Imaging
Solutions. I'll discuss the visualization process and open-source tools
that we use to render images as well as some of the image processing
components that I've developed. These selected projects were aimed at
improving image quality, enhancing automated image registration tools, and
a new feature that can synthesize planar nuclear medicine images using a
3D visualization technique. I'll finish the talk with an introduction to
image-based radionuclide dosimetry, which will be part of a new product
that I am currently designing that is aimed at monitoring and planning
radiopharmaceutical therapies that are employed for treating
neuroendocrine and metastatic prostate cancers.
By Dr. Spencer Manwell
Convergent Imaging Solutions