A novel reconstruction technique designated as Enhanced Attenuation Correction significantly reduced visual respiratory misalignment artifacts in selected positron emission tomography and computed tomography examinations, according to a retrospective evaluation published in EJNMMI Reports. In the single-center study, five experienced radiologists independently assessed coronal images from 54 patients who exhibited respiratory motion artifacts on conventional scans, with all five evaluators recording statistically significant reductions in image misalignment following the algorithmic correction.
In standard combined positron emission tomography and computed tomography (PET/CT) workflows, CT imaging is routinely utilized to calculate attenuation correction factors for reconstructing PET emission data. However, physiological respiratory motion frequently induces spatial displacement between the two acquisitions, generating respiratory misalignment artifacts near the diaphragm that have been reported to compromise up to 30 percent of lesion-based clinical evaluations and frequently require repeating both imaging acquisitions.
Prior strategies developed to mitigate respiratory displacement have included voluntary breath-holding protocols and external respiratory gating during free breathing, though both approaches have predominantly relied upon quantitative parameters such as standard uptake values and signal-to-noise ratios rather than overall visual interpretability. While recent artificial intelligence frameworks have also explored deep-learning corrections, the authors noted that such models present practical hurdles, including long computational processing times, opaque internal algorithms, and the necessity of assembling large device-specific training datasets.
Developed by GE HealthCare, the Enhanced Attenuation Correction (EAC) method directly evaluates PET artifacts from an initial standard PET volume against CT data to derive control parameters that reshape the CT volume according to modeled anatomical respiratory patterns. Because the algorithm relies exclusively on acquired patient imaging data, it operates without requiring repeat radiation exposure, additional scan time, or extensive deep-learning neural network libraries, enabling post-acquisition application in free-breathing patients.
The retrospective investigation evaluated 179 consecutive patients who underwent F-18 fluorodeoxyglucose (FDG) PET/CT imaging at the authors’ institution between July and August 2024. Following the exclusion of patients younger than 12 years, individuals exhibiting gross body movement during scanning, and those positioned non-prone, the final analyzed cohort comprised 145 examinations from 79 male and 66 female patients with a mean age of 65.1 ± 15.6 years and a median age of 69 years.
All examinations were conducted using an Omni Legend 32 scanner with PET acquisitions captured at two minutes per bed position, a 500-millimeter field of view, a 384 by 384 matrix size, and 2.07-millimeter slice thickness under Advanced MotionFree data-driven respiratory gating. Complementary CT scans were obtained at a tube voltage of 120 kilovolts, tube currents ranging between 80 and 220 milliamperes, a slice thickness of 3.75 millimeters, and slice spacing of 2.07 millimeters, followed by block sequential regularized expectation maximization reconstruction with Precision DL deep-learning processing.
To identify candidates for algorithmic correction, two nuclear medicine radiologists independently screened the conventional scans, establishing a misalignment cohort of 54 patients (36 males and 18 females, with an average age of 64.1 ± 15.0 years) where at least one reader observed diaphragmatic deviation. The researchers opted for visual assessment rather than tumor-based standard uptake values because target lesions are not invariably present; five radiologists with over three years of nuclear medicine experience independently scored side-by-side coronal slices using a subjective 1-to-5 misalignment scale ranging from no significant misalignment to severe deviation.
Statistical comparison using the Wilcoxon signed-rank test confirmed statistically significant reductions in misalignment scores across all five evaluating radiologists (p < 0.01). Prior to EAC application, mean scores across the readers were 3.5, 3.4, 3.3, 3.1, and 3.1; following correction, these averages decreased to 2.1, 2.4, 2.5, 2.2, and 1.9, respectively. Individual evaluators determined that between 63 percent and 87 percent of cases improved with EAC, while worsening was observed in only 4 percent to 11 percent of cases.
Inter-rater agreement measured through intraclass correlation coefficients (ICC(2,1)) proved relatively low, ranging from 0.13 to 0.54 before correction and 0.09 to 0.51 after correction, which the authors attributed to inherent differences in individual visual perception rather than clinical diagnostic experience. Crucially, when the authors separately inspected 91 examinations that initially exhibited no respiratory misalignment, EAC processing caused artifact deterioration in 35 cases (38.4 percent), indicating that the algorithm can occasionally misinterpret non-diaphragmatic anatomical structures when applied indiscriminately.
In light of these selective failures, the authors advised applying EAC strictly as a secondary post-processing measure when diaphragmatic misalignment is affirmatively identified on baseline reconstructions. The investigators acknowledged key study limitations, including its retrospective single-center design, the inability to verify diagnostic performance changes due to a limited number of evaluable tumor lesions, and the absence of direct comparisons against other motion-correction techniques, underscoring the need for standardized scoring systems and prospective multi-center validation.
Radiopharmaceutical administration and patient preparation adhered to specific clinical protocols within the nuclear medicine facility. Eligible patients fasted for at least four hours prior to receiving an intravenous injection of F-18 fluorodeoxyglucose administered at four megabecquerels per kilogram of body weight. Following tracer administration, an uptake interval of 60 minutes preceded scanning. During the imaging sequence, breathing proceeded freely without physician intervention or mandatory breath-holding instructions. Image reconstruction incorporated a beta value of 300 in the block sequential regularized expectation maximization algorithm, operating in conjunction with the deep-learning reconstruction platform across baseline and corrected acquisitions.
To organize the qualitative visual evaluation, the research team instituted a standardized viewing protocol. After the initial screening pair identified the 54 cases displaying diaphragmatic displacement, another nuclear medicine radiologist randomized the presentation order of the examinations. Five diagnostic radiologists, each with more than three years of experience interpreting nuclear medicine studies, reviewed the randomized series in a single sitting without repeated sessions. Side-by-side coronal sections of uncorrected baseline scans and enhanced attenuation correction reconstructions were viewed on standard high-resolution diagnostic monitors routinely employed for daily clinical PET/CT interpretation.
Statistical analyses conducted in JMP version 17 established defined interpretation tiers for inter-rater agreement, classifying intraclass correlation coefficient values below 0.50 as poor, 0.50 to 0.74 as moderate, 0.75 to 0.89 as good, and 0.90 or greater as excellent. While overall score agreement remained low, inter-rater concordance regarding whether correction produced visual improvement reached moderate levels among select observer pairs, specifically between evaluators B and D, B and E, and C and D, with agreement rates spanning 0.35 to 0.63. Moreover, score difference analyses revealed consistent downward trajectories across all five readers, with 95 percent confidence intervals spanning negative 2.0 to negative 1.0 for two evaluators, negative 1.0 to negative 1.0 for two evaluators, and negative 1.0 to 0.0 for one evaluator.
Individual case analyses underscored both the efficacy and the limits of the algorithmic adjustments on coronal PET sections. In a representative case involving a male patient in his sixties, the mean visual misalignment score dropped from 3.6 on the pre-correction baseline image to 1.0 following enhanced attenuation correction, reflecting substantial realignment along the diaphragm. In a second male patient in his sixties displaying severe baseline displacement, the average score improved from 5.0 to 3.6 after correction; notable anatomical improvement occurred, although residual misalignment remained visible. For both evaluated patients, standardized uptake value displays were maintained on a consistent scale from zero to three across baseline and corrected images to facilitate standardized visual inspection.
The authors also detailed instances where algorithmic processing failed to resolve misalignment or resulted in worsening. In a case involving a female patient in her fifties, marked diaphragmatic displacement persisted after correction, with the mean score shifting only slightly from 5.0 on the baseline image to 4.2 on the corrected image. In another scan from a male patient in his seventies, correction produced worsened misalignment, with scores recorded as 5.0 before correction and 3.6 afterward alongside visible artifact exacerbation. The investigators speculated that in non-misaligned examinations, the algorithm mistakenly identified adjacent structures resembling the diaphragm. In contrast, in certain cases with severe displacement, the software failed to identify the misaligned diaphragm, though specific predisposing conditions could not be identified.
Beyond the primary test cohort, the researchers examined broader institutional data to determine whether respiratory displacement was specific to detector hardware. In the primary group of 145 examinations on the silicon photomultiplier scanner, diaphragmatic misalignment was identified in 54 cases, representing 37.2 percent of scans. Supplemental institutional evaluations showed consistent artifact rates across different scanner architectures, with misalignment identified in 56 of 161 scans (34.8 percent) on another silicon photomultiplier unit and in 34 of 93 scans (36.6 percent) on a conventional photomultiplier tube system.
In clinical practice, the retrospective algorithm provided workflow advantages without requiring standardized technologist breathing guidance during PET/CT acquisitions. Because correction utilizes existing emission and transmission data, patients avoid prolonged table times and the additional radiation dose incurred by repeat computed tomography scans. Discussing methodological considerations, the investigators noted that conventional quantitative metrics like tissue standard deviation or coefficient of variation in the liver and lungs are problematic because the correction process alters organ homogeneity. As a result, the authors emphasized that future research should focus on standardizing visual scoring criteria, measuring physical displacement lengths or areas before and after correction, and evaluating lesion uptake values to validate enhanced attenuation correction across wider diagnostic settings.
Subject of Research: Evaluation of enhanced attenuation correction for respiratory motion artifacts in PET/CT imaging
Article Title: Enhanced attenuation correction in FDG-PET/CT with respiratory mismatch correction
Article References: Kitamura, Y., Isoda, T., Sando, M., Kusunoki, M., Kikuchi, K., Baba, S., & Ishigami, K. (2025). Enhanced attenuation correction in FDG-PET/CT with respiratory mismatch correction. EJNMMI Reports, 9(1), Article 43. https://doi.org/10.1186/s41824-025-00279-z
Image Credits: AI Generated
DOI: 10.1186/s41824-025-00279-z
Keywords: enhanced attenuation correction, respiratory misalignment artifacts, FDG-PET/CT, positron emission tomography, computed tomography, Omni Legend 32, misalignment score, nuclear medicine, radiologist visual evaluation, diaphragm motion, Advanced MotionFree, Precision DL, intraclass correlation coefficient, attenuation correction, F-18 fluorodeoxyglucose, Wilcoxon signed-rank test, respiratory gating
News Source: Ophelia Keating. (November 13, 2026). Enhanced Attenuation Correction Reduces Respiratory Misalignment in Selected PET-CT Scans. Scienmag.
