Cdhn: Cross-Domain Hallucination Network for 3d Keypoints Estimation

Zohaib, Mohammad; Padalkar, Milind G.; Morerio, Pietro; Taiana, Matteo; Bue, Alessio Del

doi:10.2139/ssrn.4349267

Cited by 1 publication

(4 citation statements)

References 49 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Global correspondence methods have focused on identifying class-specific keypoints from point clouds [14], [47]- [50]. SC3K is a recently proposed model that accurately infers sparse key points given imperfect input point clouds but struggles to predict dense output points [51]. These approaches establish correspondence but were not developed for the task of anatomical SSM, where the training sample size is typically limited, and dense points are required to capture anatomical detail.…”

Section: Related Workmentioning

confidence: 99%

“…In inference, we set the target as the point cloud with the minimum CD to all others to acquire population-level correspondence. • SC3K [51] is the Self-supervised and Coherent 3D Key- points (SC3K) estimation method proposed in concurrence with Point2SSM. This method detects sparse keypoints from point clouds that are noisy, down-sampled, and arbitrarily rotated.…”

Section: B Single Anatomy Experimentsmentioning

confidence: 99%

“…This method detects sparse keypoints from point clouds that are noisy, down-sampled, and arbitrarily rotated. • SC3K* [51] denotes the SC3K model with our Point2SSM++ loss (Eq. 5).…”

Section: B Single Anatomy Experimentsmentioning

confidence: 99%

“…5). The original SC3K loss was proposed for sparse output points (M ≈ 10) and was shown to fail given denser output (M > 50) [51]. Thus, we include SC3K* to compare the SC3K and Point2SSM++ architectures alone.…”

Section: B Single Anatomy Experimentsmentioning

confidence: 99%

See 3 more Smart Citations

Learning spatiotemporal statistical shape models for non-linear dynamic anatomies

Adams

Khan

Morris

et al. 2023

Front. Bioeng. Biotechnol.

View full text Add to dashboard Cite

Numerous clinical investigations require understanding changes in anatomical shape over time, such as in dynamic organ cycle characterization or longitudinal analyses (e.g., for disease progression). Spatiotemporal statistical shape modeling (SSM) allows for quantifying and evaluating dynamic shape variation with respect to a cohort or population of interest. Existing data-driven SSM approaches leverage information theory to capture population-level shape variations by learning correspondence-based (landmark) representations of shapes directly from data using entropy-based optimization schemes. These approaches assume sample independence and thus are unsuitable for sequential dynamic shape observations. Previous methods for adapting entropy-based SSM optimization schemes for the spatiotemporal case either utilize a cross-sectional design (ignoring within-subject correlation) or impose other limiting assumptions, such as the linearity of shape dynamics. Here, we present a principled approach to spatiotemporal SSM that relaxes these assumptions to correctly capture population-level shape variation over time. We propose to incorporate modeling the underlying time dependency into correspondence optimization via a regularized principal component polynomial regression. This approach is flexible enough to capture non-linear temporal dynamics while encoding population-specific spatial regularity. We demonstrate our method’s efficacy on synthetic data and left atrium segmented from cardiac MRI scans. Our approach better captures the population modes of variation and a statistically significant time dependency than existing methods.

show abstract

Section: Related Workmentioning

confidence: 99%

Section: B Single Anatomy Experimentsmentioning

confidence: 99%

“…This method detects sparse keypoints from point clouds that are noisy, down-sampled, and arbitrarily rotated. • SC3K* [51] denotes the SC3K model with our Point2SSM++ loss (Eq. 5).…”

Section: B Single Anatomy Experimentsmentioning

confidence: 99%

Section: B Single Anatomy Experimentsmentioning

confidence: 99%

See 2 more Smart Citations