Medical Image Reconstruction
Class meeting time: Tuesdays 6:00p-8:00p @
BME 3201 auditorium
Attendance is required for all lectures - Attending 85% of
classes is required to pass the course NO EXCEPTIONS
Intended Learning Objectives (ILOs)
To build a strong theoretical and practical
implementation skills for medical image reconstruction. This is an
advanced course and therefore will cover advanced topics immediately
from the start.
A number of sources from textbook chapters
and research papers will be handed out for each lecture
Material for mathematical background lecture 1
Handout #2: Material for mathematical background lecture 2
#3: Reference for interlaced Fourier transform
Material for Shepp-Logan phantom
Material for Partial Fourier Methods
Material for Conventional Gridding
#7: Material for Matrix Equation Gridding
Handout #8: Material for Motion Estimation
Reference for ultrasound imaging.
Reference for synthetic aperture ultrasound imaging.
Handout #11: Material for Tomography lecture.
#12: Material for Super-Resolution lecture.
100% on Class Projects
Course Topics to be Covered (Tentative)
- Mathematical basis (Matrix computations
and Fourier optics)
- Image reconstruction in CT/MRI/ultrasound imaging
- Compressed sensing theory and
applications (if time allows)
Radial sampling of the analytical Shepp-Logan phantom
file, target image size should be128x128, each row is one sample,
four numbers on each row for normalized (kx,ky) and their complex
k-space data value, format: "kx value" "ky value" "real part of
k-space" "imaginary part of k-space").
Full k-space data of a real image of transverse slice of a normal
human brain for partial Fourier reconstruction data. (Download
sample Matlab reading program here)
file, k-space size: 256x256, each row is one sample with real and
imaginary parts, 2D k-space written in sequence row by row)
Synthetic aperture ultrasound
is written in binary form - information about organization of data
is included inside each archive - choose one archive only to work
To be announced.