The aim of this project is change detection between medical images (CT scans and MRI) for later medical research and treatment, using the iterative Ricci flow algorithm
The aim of this project is change detection between medical images (CT scans and MRI) for later medical research and treatment, using the iterative Ricci flow algorithm. In this project, we are expanding the use of the Ricci flow for the two-dimensional images, and implementing the algorithm for three-dimensional model that built from a collection of two-dimensional CT and MRI scans. The main advantage of the Ricci flow over other flows applied on images, is that the Ricci flow takes into account the geometry of the input. In addition, the project is based on the assumption that the three-dimensional geometric model contains information that may not be achieved by merely looking at two-dimensional slice.
During the project, we were asked to convert the scans in DICOM format to 3D Model, using the 3D-Slicer software, and put them in as an input to MATLAB. Our algorithm runs on the input and discovers the most remarkable changes in the curvature of the model. Finally, it marks the “suspicious” areas, i.e. areas where the sharpest change of the curvature was.
System block scheme
Ricci flow
- Geometric flow that deforms the metric g(t) of a manifold
- Smoothing up for a uniform curvature
- Evolve the metric rather than the image itself
- Using Ricci curvature function Ric(g(t))
Discrete Ricci flow
The algorithm steps:
- Compute the weights of edges and faces
- Compute the Ricci curvature Ric(e) for each edge
- Apply the evolution step for edge’s lengths:
- Repeat steps 1-3 until a stopping condition is reached
Simulation results
- 3 iterations
- 300 suspicious edges
noise detection:
