3D image processing and analysis has drawn great attention to the field of medical imaging. The main goal of this course is to provide introduction to medical image computing techniques. The course focuses on voxel- and surface-based techniques including shape correspondence, surface labeling, shape quantification, etc. The course covers how the quantitative techniques can support understanding of human cognition/behaviors or brain disorders.

This course is designed for graduate students who have no prior experiences on medical image processing and shape analysis. Advanced techniques particularly on structural MRIs will be covered, such as mesh reconstruction, image segmentation/registration, graph/spherical CNNs, etc. Prior knowledge on image processing and computer graphics will be useful but not essential. This course assumes that audiences are familiar with fundamental concepts of linear algebra, probability, and multivariate calculus and comfortable with programming languages.

Attendance: 10%, Individual assignments: 40%, Final exam: 50%

Lecture slides will be provided.

[Week Contents]
1. Introduction
- Course overview
- Image modalities, FreeSurfer package
2. Image Acquisition
- Medical imaging history, ionizing imaging modalities (X-Rays/CT, PET/SPECT)
- Non-ionizing imaging modalities (MRI principles, sMRI, dwMRI, fMRI)
3. Basic Image Processing Techniques
- MR imaging issues (noise, motion, inhomogeneous magnetic field)
- Convolution theory, Fourier transform, filtering
- Contrast enhancement, simple segmentation (Otsu, EM, watershed, etc.)
- Mathematical morphology (morphological operations)
4. Discrete Surface Representation
- Surface properties (manifold, orientability, genus-#, Euler characteristics, surface normal)
- Discrete representation (level-set, polygon mesh) and data structure
- Visualization tools (ParaView, in-class tutorial)
5. Surface Reconstruction
- Basic human brain anatomy (cortical layers)
- Volumetric image processing (coordinate transformations, skull stripping, tissue mask)
- Surface reconstruction (surface evolution, marching cube)
- Topological correction and spherical mappings
6. Basic Surface Processing
- Surface retessellation (icosahedral subdivision, triangle intersection test, barycentric interpolation)
- Principal curvatures, geodesics
- Spectral representation (eigenfunctions of Laplace-Beltrami operator, spherical harmonics)
- Mesh smoothing
7. Surface Registration I
- Coordinate system in MR scanners (common space, affine transformation)
- Homeomorphism and diffeomorphism
- Mathematical formulation
8. Midterm week
9. Surface Registration II
- Optimization (Taylor series, first & second order approximation, gradients on triangles)
- Parametric registration, spherical registration
- Spectral matching, landmark matching
- Evaluation metrics (Jacobian, warping distortion, NCC, etc.)
10. Landmark Extraction
- Relation of sulcal patterns to functions & cognition
- Anatomical variability
- Surface valley (sulci) detection and labeling
11. Surface Quantification
- Anatomical biomarkers (cortical thickness, surface area, etc.)
- Gyrification & sulcal depth (Hamilton-Jacobi PDE)
- Term project proposal presentation
12. Surface Annotation I
- Cortical parcellation protocols
- Classic surface annotation (single atlas/multiatlas surface labeling)
13. Surface Annotation II
- Deep learning-based surface annotation (graph CNNs, spherical CNNs)
- Surface data augmentation techniques
- Post-processing (graph-cut)
- Evaluation metrics (Dice overlap, Jaccard overlap, etc.)
14. Statistical Shape Analysis
- Group analysis, ROI-based analysis
- Orthogonal Procrustes alignment and shape PCA
- Vertex-wise statistical analysis, mixed effects models, multi-comparison corrections
15. Course wrap-up
16. Final Exam

- Offline classes only
- Letter grade only
- Lecture, term project