Computed Tomography has become an essential first-line imaging tool for diagnosing a wide range of brain conditions, including stroke, trauma, and tumors. In particular, CT plays a critical role in the rapid assessment of acute ischemic stroke, where timely and accurate imaging is vital for treatment decisions. However, patient head motion—a common challenge in emergency settings and reported in nearly 25% of acute stroke cases—can significantly degrade image quality [1,2]. This often necessitates repeat scans, which prolong examination time and increase radiation exposure to patients. Motion Freeze, a deep learning–based algorithm, effectively reduces motion artifacts, improving neuro-imaging outcomes.

Dedicated Deep Learning Solution for Head Motion Artifact Reduction

Motion Freeze is the first deep learning algorithm on the market* specifically developed to correct head motion artifacts. Its neural network architecture was designed from the ground up for this task and trained on a large dataset generated from diverse, clinically realistic motion scenarios. By significantly reducing artifacts and improving the visibility of brain structures and lesions, Motion Freeze helps minimize the need for repeat scans—streamlining workflow and reducing radiation exposure for patients.



*FDA cleared, MDR marked


3D modeling of the motion pattern to restore the real clinical situation

Patient head movement typically involves multiple motion patterns, and no current algorithm can effectively eliminate the resulting artifacts. To establish a gold-standard dataset, the Motion Freeze algorithm simulates artifacts in the X, Y, and Z directions, including rotation, translation, oscillation, and mixed scenarios. By introducing diverse motion artifacts, this approach enables the trained network model to handle a broad spectrum of motion conditions.

Motion Freeze effectively suppresses head motion artifacts, delivering clear visualization of brain structures without obscuring lesions. By reducing the need for repeat scans, it saves time, conserves resources, and minimizes patient radiation exposure.

The algorithm supports a wide range of head scan protocols—including non-contrast, contrast-enhanced, CTA, and perfusion—improving image quality and diagnostic accuracy in neuroimaging.

Head motion artifact suppression

Motion Freeze effectively suppresses head motion artifacts, delivering clear visualization of brain structures without obscuring lesions. By reducing the need for repeat scans, it saves time, conserves resources, and minimizes patient radiation exposure.

Applicable for various protocols

The algorithm supports a wide range of head scan protocols—including non-contrast, contrast-enhanced, CTA, and perfusion—improving image quality and diagnostic accuracy in neuroimaging.

Scan:
Head Axial
1.0 sec/rotation
kV: 120
mAs: 347
CTDIvol: 38.7 mGy
Eff.Dose: 1.5 mSv

Contrast:
-

Recon:
H_SOFT_B, KARL:5
5.0 mm * 5.0mm
WW/WL: 80/35

Scanner:
uCT 960+

Scan:
Head Axial
1.0 sec/rotation
kV: 120
mAs: 347
Eff.Dose: 1.6 mSv

kV: 120
mAs: 347
Eff.Dose: 1.6 mSv

Contrast:
-

Recon:
H_SOFT_B, KARL:5
5.0 mm * 5.0mm
WW/WL: 80/35

Scanner:
uCT 960+

Scan:
Head Axial
0.5 sec/rotation
kV: 120
mAs: 247
CTDIvol: 38.7 mGy
Eff.Dose: 1.3 mSv

Contrast:
-

Recon:
Matrix: 512 * 512
Slice: 5.5 * 5.5 mm
HIR: H_SOFT_B/KARL: 3

Scanner:
uCT 868

Scan:
Head Axial
1.0 sec/rotation
kV: 120
mAs: 346
Eff.Dose: 1.59 mSv

Contrast:
-

Recon:
Matrix: 512 * 512
Slice: 5.5 * 5.5 mm
HIR: H_SOFT_B/KARL: 3

Scanner:
uCT 960+

Scan:
Head Axial
0.5 sec/rotation
kV: 100
mAs: 100
CTDIvol: 188.2 mGy
Eff.Dose: 6.3 mSv

Contrast:
55 mL, 5.5 mL/s

Recon:
HIR:
H_VSOFT_A, KARL:8
5.0 mm * 5.0mm
WW/WL: 80/35
AIIR:
HeadStandard/AIIR: 2

Scanner:
uCT 960+

Scan:
Head Axial
1.0 sec/rotation
kV: 120
mAs: 347
Eff.Dose: 1.5 mSv

kV: 120
mAs: 297
Eff.Dose: 1.5 mSv

Contrast:
55 mL, 5.5 mL/s

Recon:
HIR:
H_VSOFT_A, KARL:8
5.0 mm * 5.0mm
WW/WL: 80/35
AIIR:
HeadStandard/AIIR 2

Scanner:
uCT 960+

The AI-based motion correction (Motion Freeze) algorithm was evaluated in 53 cerebral CT cases with motion artifacts that required immediate rescans. First-scan images were reconstructed with both standard iterative reconstruction (IR) and the MF algorithm, while rescanned images served as reference. Compared to motion-affected images, MF significantly improved image quality—raising SNR and CNR, reducing MSE by 44.1%, and enhancing PSNR, SSIM, and MI (all p < 0.001). Subjective scores and lesion detectability were also higher, with ASPECTS AUC increasing from 0.614 to 0.817. The study confirms that MF effectively reduces motion artifacts and enhances diagnostic performance in cerebral CT.

< Clinical validation of an AI-based motion correction reconstruction algorithm in cerebral CT Eur Radiol. 2022 Dec;32(12):8550-8559.>.

—《European Radiology 》

Motion Freeze — An innovatively designed 3D neural network for suppressing head motion artifacts, was evaluated against other neural networks and clinically tested on 30 artifact-affected image sets. It outperformed all comparators, achieving the lowest normalized root-mean-square error (NRMSE) and the highest peak signal-to-noise ratio (PSNR) and structural similarity index (SSIM), demonstrating superior artifact reduction and image quality.

< A deep learning method for eliminating head motion artifacts in computed tomography 2022 Jan;49(1):411-419. >.

—《Medical Physics. 》

1. Fahmi F, Beenen LF, Streekstra GJ, Janssen NY, de Jong HW, Riordan A, Roos YB, Majoie CB, Vanbavel E, Marquering HA. Head movement during CT brain perfusion acquisition of patients with suspected acute ischemic stroke. Eur J Radiol. 2013 Dec;82(12):2334-41. doi: 10.1016/j.ejrad.2013.08.039. Epub 2013 Aug 30. PMID: 24041432.Hernandez, D., Eldib, M. E., Hegazy, M. A., Cho, M. H., Cho, M. H., & Lee, S. Y. (2018). A head motion estimation algorithm for motion artifact correction in dental CT imaging. Physics in Medicine & Biology, 63(6), 065014.
2. Moghari, Mahdieh Dashtbani, et al. "Head Movement During Cerebral CT Perfusion Imaging of Acute Ischaemic Stroke: Characterisation and Correlation With Patient Baseline Features." European Journal of Radiology, vol. 144, Sept. 2021, p. 109979

Learn more about Motion Freeze: Motion Freeze Technical White Paper