Spaces:
Sleeping
Sleeping
| title: Image Registration Demo | |
| emoji: 🔄 | |
| colorFrom: blue | |
| colorTo: green | |
| sdk: docker | |
| pinned: false | |
| license: mit | |
| # image-registration | |
| This is a demonstration of how transformation matrices affect registration for the affine case. | |
| [Click here to run this on Streamlit](https://tinyurl.com/registration-demo). | |
| ## What is Registration? | |
| Image Registration aims to align the images to a common reference space (in the medical image case, typically named atlas) | |
| * Given: reference/fixed image A (atlas) and floating/moving image B. | |
| * Task: find transformation T (for 2D images, a 3x3 matrix), such that T(B) is similar to A | |
| * Similarity: defined through a similarity measure C | |
| ## What are typical reference and floating images? | |
| Types of reference image A: | |
| * Inter-subject registration to an atlas, i.e., a population-based image | |
| * Inter-subject registration to a different subject | |
| * Intra-subject registration, e.g., T1-weighted to T2-weighted image registration of same subject | |
| ## What are some challenges with Registration? | |
| * Non-rigid tissues (e.g., gray/white matter or fatty tissue) | |
| * Pre-/Post-operative (e.g., blood, resection cavities) | |
| * Imaging artifacts (e.g., Motion, Metallic Transplant) | |
| ## What is the difference between Rigid and Affine transforms? | |
| Rigid registration allows for altering the roll, pitch and yaw of the object to move it in space but the size and shape of the object cannot change. | |
| On the other hand, an affine registration allows altering the scale/size of the object and introduce shear (where straight lines remain parallel). Affine is more general than Rigid. |