Overcoming Through Plane Motion in Dynamic Imaging

98 concepts78 papers1 proposal

Imagine filming a dolphin jumping out of the ocean using a camera fixed strictly at the water's surface, looking straight down. As the dolphin breaks the surface, it appears in your 2D frame as an expanding and contracting shape, then vanishes entirely. To a 2D tracking algorithm, this object is deforming unphysically, violating mass conservation and continuity.

This scenario illustrates the 2D-to-3D dimensionality gap in medical and biological imaging. When dynamic, three-dimensional structures are captured via standard 2D slice sequences—such as cardiac cine-MRI, ultrasound, or planar video—moving tissue frequently travels perpendicular to the imaging plane. This phenomenon, known as through-plane or out-of-plane motion, introduces severe tracking inaccuracies and artifacts because the true physical motion cannot be fully captured within a single planar coordinate system.

Why 2D Tracking Breaks Down

Traditional image registration and tracking algorithms assume that the intensity, texture, or features of a structure remain consistent across frames. When tissue moves out of the imaging plane, this assumption fails completely.

In ultrasound imaging, as noted by María J. Ledesma‐Carbayo et al. in Spatio-temporal nonrigid registration for ultrasound cardiac motion estimation, through-plane movement causes speckle decorrelation. The acoustic signature changes not because the tissue deformed within the plane, but because entirely new tissue entered the slice.

Similarly, in quantitative cardiac MRI mapping (such as $T_1$ or $T_2$ relaxation mapping), respiratory and cardiac cycles push anatomy across slice boundaries. As Sébastien Roujol et al. demonstrate in Impact of motion correction on reproducibility and spatial variability of quantitative myocardial T2 mapping, standard 2D registration frameworks are fundamentally limited because they only correct for in-plane adjustments. Any tissue moving out of the slice results in a permanent loss of consistent anatomical information across the image series. Neel Dey in Data Consistent Deep Rigid MRI Motion Correction reinforces this, explaining that failing to account for these three-dimensional and time-varying effects prevents retrospective techniques from effectively resolving real-world clinical artifacts.

The Clinical and Biological Blind Spots

The inability to track out-of-plane motion creates major diagnostic blind spots across multiple domains:

  • Cardiology: The human heart undergoes a complex, volumetric "wringing" motion involving simultaneous twisting, longitudinal shortening, and radial thickening. As argued by Vithu Logan Jeya in Diagnosing Cardiac Deformations using 3d Optical Flow, 2D techniques fail to capture this full volumetric complexity, obscuring critical indicators of cardiac health that only manifest in three dimensions.
  • Speech and Neuromechanics: In speech therapy, Yin Yang 0002 demonstrates in Contour-based 3D tongue motion visualization using ultrasound image sequences that relying on a single mid-sagittal ultrasound slice completely ignores lateral expansion and twisting of the tongue muscle. Similarly, tracking muscle units via 2D ultrasound cross-sections fails to simultaneously resolve longitudinal shortening and radial expansion, a limitation highlighted in comparative studies of ultrafast ultrasound images.
  • Behavioral Analysis: Even in macro-scale video analysis, such as assessing infant neurodevelopment, projecting complex 3D body movements onto 2D video coordinates discards crucial depth-dependent coordination and subtle rotational nuances (Binh Nguyen-Thai et al. in A Spatio-Temporal Attention-Based Model for Infant Movement Assessment From Videos).

Multi-View Fusion: Intersecting Planes

How do researchers overcome this without requiring full, high-resolution 3D real-time scans, which are often clinically unfeasible due to hardware constraints or low through-plane resolution?

One powerful approach is Multi-View Fusion. Instead of relying on a single slice orientation, frameworks combine information from orthogonal planes. For instance, Ismail Ben Ayed et al. developed a track-to-track fusion method that integrates short-axis and long-axis cardiac MRI views, allowing motion perpendicular to one plane to be captured by the other.

Building on this concept, Qingjie Meng et al. introduced MulViMotion: Shape-Aware 3D Myocardial Motion Tracking From Multi-View Cardiac MRI. This deep learning architecture uses a shape-aware regularization module to fuse multi-view 2D slices into an anatomically consistent 3D deformation field, effectively bypassing the low longitudinal resolution of standard clinical scans.

[Short-Axis 2D Slice]  ──┐
                         ├─► [Shape-Aware Regularization] ─► True 3D Deformation Field
[Long-Axis 2D Slice]   ──┘

Physics-Grounded and Mesh-Based Priors

Another paradigm shifts the tracking target from unconstrained pixel grids to structured geometric models governed by physical laws. By embedding a biomechanical or geometric prior into the tracking pipeline, the system can infer what happens between or outside the imaging planes.

Early foundations by Pengcheng Shi et al. in A Model-Based Integrated Approach to Track Myocardial Deformation Using Displacement and Velocity Constraints anchored sparse 2D data points within a physics-based biomechanical model to reconstruct dense 3D maps. Modern iterations leverage deep generative models to achieve similar constraints. For example, Chen Qin et al. developed a framework that learns a latent space of physically plausible 3D heart deformations, allowing the model to robustly account for through-plane motion during inference by exploring a biomechanics-informed prior space.

To bridge the gap between discrete 2D slices and continuous 3D geometry, Qingjie Meng et al. proposed a Mesh-based 3D Motion Tracking in Cardiac MRI using Deep Learning framework. Instead of tracking voxels, they model the organ as a 3D geometric mesh and employ a differentiable mesh-to-image rasterizer. This allows the network to learn directly from standard 2D slices while optimizing a fully realized 3D deformation model, resolving the out-of-plane tracking problem by design.

Key Foundational Work

A Model-Based Integrated Approach to Track Myocardial Deformation Using Displacement and Velocity Constraints (1995) — contributed 2 concept elements to this direction.
History & citationVersion 2 · 2026-10-04

Current · Permanent link

  1. Version 2 (latest) · 2026-10-04 · 74 papers
  2. Version 1 · 2026-10-04 · 107 papers
Cite this direction

Latest version (a living page):

@misc{lacuna_out_of_plane_motion_artifacts_ed62d12b,
  title = {out-of-plane motion artifacts},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/out-of-plane-motion-artifacts/rdr_ed62d12bc0e648499f5ae835e647986b},
  note = {Latest version of a living research direction; cite a version for a fixed text. Accessed 2026-11-19},
}

This exact version (version 2, frozen text):

@misc{lacuna_out_of_plane_motion_artifacts_ed62d12b_v2,
  title = {out-of-plane motion artifacts},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/out-of-plane-motion-artifacts/rdr_ed62d12bc0e648499f5ae835e647986b/v/2},
  note = {Version 2, frozen 2026-10-04. Accessed 2026-11-19},
}
About

These directions overlap with this one through shared papers.

See something wrong? Report a mistake