# Cardiac device motion tracking from kilovoltage projections during stereotactic arrhythmia radioablation

## TL;DR

This study develops a method to track motion of cardiac devices during radioablation, achieving high accuracy in motion estimation.

## Contribution

A novel motion tracking method for cardiac radioablation using kilovoltage projections and ICD lead tip segmentation.

## Key findings

- ICD lead tip segmentation achieved 85% accuracy with positions within 2 mm of ground truth.
- Three-dimensional motion estimation had errors < 2 mm in 78% of cases.
- Median motion displacements were 1.4 mm (left-right), 1.5 mm (superior-inferior), and 0.8 mm (anterior-posterior).

## Abstract

•Cardiac and respiratory motion tracking method for cardiac radioablation.•Implantable cardioverter-defibrillator lead tip segmentation with 85% accuracy.•Three-dimensional motion estimation with errors < 2 mm in 78% of cases.•Left–right, superior–inferior, and anterior–posterior motions: 7.4, 10.4, and 6.3 mm.

Cardiac and respiratory motion tracking method for cardiac radioablation.

Implantable cardioverter-defibrillator lead tip segmentation with 85% accuracy.

Three-dimensional motion estimation with errors < 2 mm in 78% of cases.

Left–right, superior–inferior, and anterior–posterior motions: 7.4, 10.4, and 6.3 mm.

Cardiac and respiratory motion present a challenge to accurate dose delivery during stereotactic arrhythmia radioablation (STAR). This study investigated a motion tracking method using kilovoltage (kV) images on a gantry‐mounted linear accelerator during STAR.

An implantable cardioverter‐defibrillator (ICD) lead tip was used for motion tracking. Planning computed tomography and cone beam computed tomography scans were analysed retrospectively on ten patients treated with STAR in RAdiosurgery for VENtricular TAchycardia (RAVENTA) (four patients underwent deep inspiratory breath‐hold, two with free breathing, three under abdominal compression, and one using high frequency ventilation). A template‐matching approach was developed to segment ICD lead tips in kV projections, with successful segmentation defined as a segmented position within 2 mm of the two‐dimensional (2D) ground truth derived from manual annotations. A probability distribution function method was used to convert 2D coordinates to three‐dimensional (3D) trajectories. 3D trajectories from the 2D ground truth were used to measure ICD lead tip displacements from its mean position. Differences between 3D trajectories from 2D ground truth and segmentations were analysed.

The median (range) 2D segmentation success rate was 85% (71–96%). Median (range) motion displacements were 1.4 (0.1–7.4) mm (left‐right), 1.5 (0.1–10.4) mm (superior‐inferior), and 0.8 (0.0–6.3) mm (anterior‐posterior). The median (range) proportion of 3D estimations with errors below 2 mm in all axes was 78% (43–96%).

ICD lead tip motion showed patient‐specific variability. This study presented a potential 3D motion tracking method for ICD lead tips during STAR.