Purpose: The Positron Emission Tomography (PET)/Magnetic Resonance Imaging (MRI) scanner combines two diagnostic imaging modalities, providing information on anatomy and physiology. Beneficial diagnosis areas are epilepsy ([<InlineEquation ID="IEq1"> <EquationSource Format="TEX">\(^{18}\)</EquationSource> </InlineEquation>F]Fluorodeoxyglucose (FDG)) and cancer recurrence ([<InlineEquation ID="IEq2"> <EquationSource Format="TEX">\(^{11}\)</EquationSource> </InlineEquation>C]methionine (MET)), where subject motion during PET acquisition reduces image quality, potentially compromising diagnostic accuracy. This project aimed to evaluate the impact of PET data-driven motion correction (ddMC) of these clinical PET radiotracers and to assess, for the first time, whether the automatic motion categorization reflects motion levels impacting the image quality. Methods: Eighty-nine PET scans (66 [<InlineEquation ID="IEq3"> <EquationSource Format="TEX">\(^{11}\)</EquationSource> </InlineEquation>C]MET, 23 [<InlineEquation ID="IEq4"> <EquationSource Format="TEX">\(^{18}\)</EquationSource> </InlineEquation>F]FDG) were reconstructed with ddMC and without motion correction (noMC) using the research software lmDuetto toolbox (GE Healthcare, Chicago, IL, USA), and were automatically categorized into motion groups. MRI images were segmented, and the regions of interest (ROIs) transferred to the PET space. The effect of ddMC was analyzed by relative signal differences between ddMC and noMC. Motion estimation and categorization were evaluated by normalized cross correlation (XC) over time and the proposed cumulative displacement-time histogram (cDTH). Results: Overall, ddMC increased signal values within cortical ROIs compared to noMC. In the high motion category, median relative mean signal differences were 0.61% (0.41−0.80%) for [<InlineEquation ID="IEq5"> <EquationSource Format="TEX">\(^{18}\)</EquationSource> </InlineEquation>F]FDG and 0.70% (0.61−0.79%) for [<InlineEquation ID="IEq6"> <EquationSource Format="TEX">\(^{11}\)</EquationSource> </InlineEquation>C]MET. The XC improved ([<InlineEquation ID="IEq7"> <EquationSource Format="TEX">\(^{18}\)</EquationSource> </InlineEquation>F]FDG: 0.80 to 0.97, [<InlineEquation ID="IEq8"> <EquationSource Format="TEX">\(^{11}\)</EquationSource> </InlineEquation>C]MET: 0.85 to 0.98). Low and medium motion groups had lesser impact, indicating motion correction is most relevant for high motion. The XC and cDTH identified subjects whose motion classification should be revised. Conclusion: In conclusion, the results confirm previous findings with ddMC using [<InlineEquation ID="IEq9"> <EquationSource Format="TEX">\(^{18}\)</EquationSource> </InlineEquation>F]FDG and demonstrate its suitability for lower-accumulating [<InlineEquation ID="IEq10"> <EquationSource Format="TEX">\(^{11}\)</EquationSource> </InlineEquation>C]MET. The automatic motion categorization needs re-evaluation to better reflect motion affecting PET image quality.

The impact of data driven motion correction for clinical brain PET/MRI radiotracers

期刊论文
出版模式：

开放获取
发表日期：

2026年4月15日

查看全文

Martin Bolin,
Anna Falk Delgado,
Emilia Palmér

摘要

Purpose:

The Positron Emission Tomography (PET)/Magnetic Resonance Imaging (MRI) scanner combines two diagnostic imaging modalities, providing information on anatomy and physiology. Beneficial diagnosis areas are epilepsy ([ \(^{18}\) F]Fluorodeoxyglucose (FDG)) and cancer recurrence ([ \(^{11}\) C]methionine (MET)), where subject motion during PET acquisition reduces image quality, potentially compromising diagnostic accuracy. This project aimed to evaluate the impact of PET data-driven motion correction (ddMC) of these clinical PET radiotracers and to assess, for the first time, whether the automatic motion categorization reflects motion levels impacting the image quality.

Methods:

Eighty-nine PET scans (66 [ \(^{11}\) C]MET, 23 [ \(^{18}\) F]FDG) were reconstructed with ddMC and without motion correction (noMC) using the research software lmDuetto toolbox (GE Healthcare, Chicago, IL, USA), and were automatically categorized into motion groups. MRI images were segmented, and the regions of interest (ROIs) transferred to the PET space. The effect of ddMC was analyzed by relative signal differences between ddMC and noMC. Motion estimation and categorization were evaluated by normalized cross correlation (XC) over time and the proposed cumulative displacement-time histogram (cDTH).

Results:

Overall, ddMC increased signal values within cortical ROIs compared to noMC. In the high motion category, median relative mean signal differences were 0.61% (0.41−0.80%) for [ \(^{18}\) F]FDG and 0.70% (0.61−0.79%) for [ \(^{11}\) C]MET. The XC improved ([ \(^{18}\) F]FDG: 0.80 to 0.97, [ \(^{11}\) C]MET: 0.85 to 0.98). Low and medium motion groups had lesser impact, indicating motion correction is most relevant for high motion. The XC and cDTH identified subjects whose motion classification should be revised.

Conclusion:

In conclusion, the results confirm previous findings with ddMC using [ \(^{18}\) F]FDG and demonstrate its suitability for lower-accumulating [ \(^{11}\) C]MET. The automatic motion categorization needs re-evaluation to better reflect motion affecting PET image quality.