# Data-Driven Motion Correction Algorithm: Validation in [13N]NH3 Dynamic PET/CT Scans

## TL;DR

A new automated motion correction algorithm was validated for dynamic PET/CT scans, showing similar results to manual methods and reducing processing time.

## Contribution

The study validates a data-driven motion correction algorithm for dynamic [13N]NH3 PET/CT and demonstrates its effectiveness in reducing observer variability.

## Key findings

- Motion correction significantly increased stress-MBF and CFR values in both patient subgroups.
- High-intensity motion was predominantly observed in the cranio-caudal direction during stress scans.
- Stress-MBF values decreased as motion extent increased, indicating motion's negative impact on quantification.

## Abstract

Background: Motion is a long-standing problem in cardiac PET/CT. An automated data-driven motion correction (DDMC) algorithm for within-reconstruction motion correction (MC) has been developed and validated in static images from [13N]NH3 and 82Rb PET/CT. This study aims to validate DDMC in dynamic [13N]NH3 PET/CT, and to explore the added value of DDMC in the evaluation of myocardial motion. Methods: Thirty-six PET/CT studies from normal patients and forty-three scans from patients with myocardial ischemia were processed using QPET software without MC (NMC), using manual in-software MC (ISMC), and DDMC. Differences in the mean values of rest-, stress-MBF, and CFR; and differences in effect size related to the use and type of MC method were explored. Moreover, motion vectors provided by DDMC were analyzed to evaluate differences in myocardial motion between scan phases and axes, and to elucidate changes in MBF quantification in relation to the motion extent. Results: In both subgroups, repeated measures ANOVA showed that the use of MC significantly increased regional and global stress-MBF and CFR values (p < 0.05), regardless of the MC method. Paired t-test analysis demonstrated a comparable ES between MC tools, despite minor differences in Cx, RCA and global rest-MBF values. High-intensity motion (>6 mm) proved to be present almost exclusively in the Z (cranio-caudal) direction. In the same axis, motion was significantly higher during stress than rest, regardless of patients’ subgroup. Finally, the Jonckheere trend test showed a significant trend caused by motion in s-MBF values, in which lower stress-MBF values were observed in response to motion extent increments. Conclusions: DDMC is feasible to perform in [13N]NH3 dynamic acquisitions and provides similar MBF/CFR values than manual ISMC. The use of DDMC reduces post-processing times and observer variability, and allows a more extensive evaluation of motion. MC is highly recommended when using QPET, as motion in the Z-axis during stress scans negatively impacts stress-MBF quantification.