Real-Time Adaptive Radiotherapy for Dynamic Anatomical Motion

You are reading version 1 of this direction, frozen 2026-10-04. The latest version is version 2.

Version 1 of 2104 papers184 conceptsfrozen 2026-10-04

Traditional radiation therapy relies on a "frozen patient" assumption. A clinician takes a high-resolution CT scan, spends days meticulously contouring the tumor and nearby organs, and designs a complex beam arrangement to deliver a precise dose. However, the moment the patient breathes or loses weight, that static plan begins to degrade.

The research direction emerging from these papers focuses on Adaptive Radiotherapy (ART): a shift from static planning to a dynamic, closed-loop system that senses, predicts, and compensates for anatomical changes in real-time.

The Two Scales of Anatomical Drift

To understand why adaptation is necessary, we must distinguish between two types of motion that invalidate static plans.

Intra-fractional motion occurs while the radiation beam is active. As noted by Xiaofei Huang and colleagues, respiratory cycles cause tumors in the lungs or liver to shift by tens of millimeters. This creates the "interplay effect," where the motion of the machine's internal shutters (collimators) synchronizes poorly with the patient's breathing, leading to localized "hot" or "cold" spots in the dose Tsukasa Wakayama.

Inter-fractional motion occurs over the weeks of a typical treatment course. Huadong Wang highlights that organs like the esophagus or bladder change shape and position daily. Furthermore, as Anqi Fu and Lei Xing point out, the tumor itself may shrink or change its sensitivity to radiation, making the initial plan increasingly obsolete.

Sensing the Change

The first step in adaptation is seeing the anatomy as it exists right now.

Modern approaches move beyond the single planning CT. Huadong Wang demonstrates the use of MR-Linac systems, which combine a linear accelerator with an MRI scanner to provide high-contrast soft tissue imaging during treatment. To make this data actionable, researchers are developing AI-driven "autosegmentation" to redraw the boundaries of the tumor in minutes rather than hours.

For faster, more frequent updates, wanli zhu and others explore combining Electronic Portal Imaging Devices (EPID) with fan-beam CT to verify the dose actually received by the patient. When high-quality imaging isn't available, "digital twins" can fill the gap. Jorge Tapias Gomez and William P. Segars developed frameworks that use mathematical models of peristalsis and breathing to simulate 4D organ deformation from a single 3D scan.

Predicting the Future

Because there is always a lag between sensing motion and adjusting the machine, the system must anticipate where the tumor will be in the next few hundred milliseconds.

Lijuan Shi and colleagues have pioneered the use of Multi-Scale Temporal Convolutional Networks (TCNs) to track thoracic movement. By treating respiratory signals as a time-series problem, these models can forecast tumor position more accurately than traditional methods.

Taking this further, Yuxiang Lai and Xiaofeng Yang introduced Auto-RMP, which treats organ motion like a natural language processing task. By "tokenizing" 4D CT scans, the system learns the "grammar" of how a specific patient's anatomy moves, allowing for tighter radiation margins and better protection of healthy tissue.

Closing the Loop: Actuation and Re-optimization

Once the motion is sensed and predicted, the treatment must change. This happens through two primary mechanisms:

  1. Mechanical Compensation: Techniques like "couch-tracking" involve moving the entire treatment table in real-time to cancel out the patient's breathing. Alexander Jöhl showed that patients habituate to this motion quickly, making it a viable clinical strategy.
  2. Algorithmic Re-optimization: Instead of just moving the patient, the system can re-calculate the radiation beams. Mark Bangert and Philipp Hennig proposed probabilistic frameworks to account for these "known unknowns" during planning.

However, adaptation is not without risk. Velibor V. Mišić discovered a "dose-reactive paradox": simply trying to "fix" yesterday's dose error today can lead to unstable fluctuations in treatment. The most effective strategies, therefore, are those that update the underlying model of the patient's physiology rather than just reacting to the last error.

Why It Works: The Intuition of "Dose Warping"

The core intuition behind this research is dose warping. In a static world, a dose of 2 Gray (Gy) is delivered to a fixed coordinate (x, y, z). In the adaptive world, we recognize that the tissue at (x, y, z) at 10:00 AM might be at (x+5, y, z) at 10:01 AM.

By using deformable image registration and real-time tracking, researchers can "warp" the planned dose onto the moving anatomy. This ensures that the energy follows the mass. As Efstratios Gavves and Jan-Jakob Sonke demonstrate, physics-guided deep learning can now perform these complex calculations in under a second, finally making the "Sense-Adapt-Deliver" cycle fast enough for daily clinical use.

Papers in this version

Showing 60 of 104 papers.

History & citationVersion 2 · 2026-10-04

Current · Permanent link

  1. Version 2 (latest) · 2026-10-04 · 89 papers
  2. Version 1 · 2026-10-04 · 104 papers
Cite this direction

Latest version (a living page):

@misc{lacuna_respiratory_motion_uncertainty_3656b1d9,
  title = {respiratory motion uncertainty},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/respiratory-motion-uncertainty/rdr_3656b1d9465a4e42ae21ceb87e97defb},
  note = {Latest version of a living research direction; cite a version for a fixed text. Accessed 2026-11-19},
}

This exact version (version 1, frozen text):

@misc{lacuna_real_time_adaptive_radiotherapy_for_dyna_3656b1d9_v1,
  title = {Real-Time Adaptive Radiotherapy for Dynamic Anatomical Motion},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/real-time-adaptive-radiotherapy-for-dynamic-anatomical-motion/rdr_3656b1d9465a4e42ae21ceb87e97defb/v/1},
  note = {Version 1, frozen 2026-10-04. Accessed 2026-11-19},
}