Free Surfer Future Directions surfer nmr mgh harvard





























- Slides: 29

Free. Surfer: Future Directions surfer. nmr. mgh. harvard. edu

Talk Outline 1. Combined Volume and Surface morphing (CVS). 2. Shape analysis. 3. Optimization of sequences for morphometry, both longitudinal and cross-sectional (with Andre van der Kouwe). 4. Other things I won’t talk about: - Integration of diffusion information (with Lilla Zollei, Anastasia Yendiki, Dennis Jen, Tim Behrens and Saad Jbabdi). - New tools for retinotopy analysis (with Eric Schwartz, Doug Greve and Jon Polimeni). - Structure/function relationships and enhanced spherical morphing (with Lilla Zollei, Doug Greve, Mert Sabuncu and Nancy Kanwisher - Inferring cytoarchitectural boundaries in vivo (with Koen Van Leemput, Jean Augustinack and Thomas Yeo). - Resting state/correlation analysis (with Doug Greve)

Combined Volume and Surface Registration Surface-based (2 D) registration does an excellent job of aligning cortical folds, but doesn’t apply to non-cortical structures (e. g. basal ganglia). Volumetric (3 D) registration applies to the entire brain but doesn’t in general align folding patterns. Solution: integrate them (5 D!). Joint work with Gheorghe Postelnicu and Lilla Zollei

Why Aligning Folds in the Volume is Hard! Affine transform of surfaces from one subject mapped to another.

5 D Registration: Example. Source Joint work with Gheorghe Postelnicu Target

5 D Registration: Preliminary Results. Target Joint work with Gheorghe Postelnicu Target

5 D Registration: Preliminary Results. 3 D (HAMMER) Joint work with Gheorghe Postelnicu 5 D

CVS Registration: Accuracy Dice Coefficients (overlap) for 10 manually labeled datasets. Joint work with Gheorghe Postelnicu

Talk Outline 1. Combines Volume and Surface analysis (CVS). 2. Shape analysis. 3. Inferring cytoarchitectural boundaries in vivo.

Shape Analysis: Spherical Harmonics (joint work with Peng Yu and Xiao Han) Reconstructed cortical surfaces using SPHARM coefficients truncated at degree 1, 2, 5, 10, 20 and the original surface

Shape Analysis: Spherical Wavelets Aj-1 Aj Aj-1 A 1 … … Bj Bj-1 Bj wavelet coefficients Wavelet decomposition of cortical surface

Shape Analysis original (synthesized) SPHARM reconstruction Wavelet reconstruction Comparison of SPHARM (left) and wavelet (right) reconstructions with largest 200 coefficients.

Results Shape Variation in Normal Population using PCA Dataset: 84 nondemented older participants (OP; 42 women: 67– 95, mean age = 80, standard derivation = 7. 25; 42 men: 71 -94, mean age = 79, standard derivation = 7. 17). In collaboration with Randy Buckner

Shape Variation in Normal Population using PCA Distinct pattern of shape variation detected with PCA Synthetic surface demonstrating Corresponding real patterns of shape variation at the cortical surface large spatial scale (yellow is largest variation)

Shape Variation in Normal Population using PCA Distinct pattern of shape variation detected with PCA Synthetic surface demonstrating Corresponding real patterns of shape variation at the cortical surface large spatial scale (yellow is largest variation)

Wavelet Shape Analysis: Newborn Growth Model Dataset: Eight normal neonates with corrected gestational ages (c. GA) of 31. 1 - 39. 72, and weeks, and 3 babies 6 months- 2 years. Built a growth model of the newborns’ cortical surface using Gompertz functions in the wavelet domain In collaboration with Peng Yu and Ellen Grant

Wavelet Shape Analysis: newborns Joint work with Peng Yu, Evelina Busa, Ellen Grant and Rudolph Pienaar

Talk Outline 1. Shape analysis. 2. Increasing sensitivity and reliability. 3. A new morphometry protocol.

Bandwidth, SNR and Distortion Problem: there is a tradeoff between distortion and SNR. - High SNR sequences (low bandwidth) have more distortion (and it only gets worse at higher field). - High bandwidth sequences have low distortion, but suffer from lower SNR. Solution: use (ridiculously) high bandwidth (≈700 Hz/voxel) and read continuously to generate multiple echoes. Bonus: T 2* information, and fieldmaps! Joint work with Anders Dale and Andre van der Kouwe

Multi-echo FLASH Sequence Advantages: • Less B 0 distortion due to its high bandwidth, and less intensity variation • Enables estimation of tissue parameters (T 1, PD, T 2*) • High CNR for more accurate subcortical segmentation Joint work with Anders Dale, Xiao Han and Andre van der Kouwe

Motivation MEF has significantly higher CNR in subcortical structures (e. g. pallidum/putamen) than MP-RAGE. BUT… MPRAGE has higher CNR for cortical gray/white segmentation. Could use MP-RAGE in addition, but it is differentially distorted with respect to the MEF scans. Solution: multi-echo MP-RAGE! Bonus: can identify dura from T 2* decay in MP-RAGE. Joint work with Anders Dale and Andre van der Kouwe

BW R/O matched protocol: MPRAGE

BW R/O matched protocol: MEMMPRAGE (RMS)

BW R/O matched protocol: MEF 30 o (mean)

BW R/O matched protocol: MEF 5 o (mean)

BW R/O matched protocol: T 2 -SPACE

New Morphometry Protocol: Identifying Dura *joint work with Andre van der Kouwe and Brad Dickerson

Morphometry Protocol PD T 1 MPRAGE T 2* TAQ≈35 min Joint work with Andre van der Kouwe, Anders Dale and Xiao Han T 2

Acknowledgements MGH Jenni Pacheco Nick Schmansky Brian T Quinn Andre van der Kouwe Doug Greve David Salat Evelina Busa Lilla Zollei Koen Van Leemput Xiao Han Kevin Teich Niranjini Rajendran Dennis Jen NINDS MGH Allison Stevens Jean Augustinack Gheorghe Postelnicu Anastasia Yendiki Neda Bernasconi Ellen Grant MIT Polina Golland Thomas Yeo Peng Yu Mert Sabuncu UC San Diego Anders Dale UCL Marty Sereno