Interscan Variability Research Articles

Diffusion MRI with Machine Learning

Abstract Diffusion-weighted magnetic resonance imaging (dMRI) of the brain offers unique capabilities including noninvasive probing of tissue microstructure and structural connectivity. It is widely used for clinical assessment of disease and injury, and for neuroscience research. Analyzing the dMRI data to extract useful information for medical and scientific purposes can be challenging. The dMRI measurements may suffer from strong noise and artifacts, and may exhibit high inter-session and inter-scanner variability in the data, as well as inter-subject heterogeneity in brain structure. Moreover, the relationship between measurements and the phenomena of interest can be highly complex. Recent years have witnessed increasing use of machine learning methods for dMRI analysis. This manuscript aims to assess these efforts, with a focus on methods that have addressed data preprocessing and harmonization, microstructure mapping, tractography, and white matter tract analysis. We study the main findings, strengths, and weaknesses of the existing methods and suggest topics for future research. We find that machine learning may be exceptionally suited to tackle some of the difficult tasks in dMRI analysis. However, for this to happen, several shortcomings of existing methods and critical unresolved issues need to be addressed. There is a pressing need to improve evaluation practices, to increase the availability of rich training datasets and validation benchmarks, as well as model generalizability, reliability, and explainability concerns.

A support vector machine-based approach to guide the selection of a pseudo-reference region for brain PET quantification.

A Support Vector Machine (SVM) based approach was developed to identify a pseudo-reference region for brain PET scans with the aim of reducing interscan and intersubject variability. By training a binary linear SVM classifier with PET datasets from two different groups, potential pseudo-reference regions were identified by considering their regional average or total contribution to the classification score. This approach was evaluated in three cohorts with different brain PET tracers: (1) 11C-PiB PET scans of Alzheimer's disease (AD) patients and age-matched controls (OC); (2) baseline and blocking scans of an 11C-UCB-J PET occupancy study; and (3) 18F-DPA-714 PET scans for healthy controls (HC) and chemo-treated women with breast cancer (BC). In the first cohort, cerebellum, brainstem, and subcortical white matter were confirmed as pseudo-reference regions. The same regions were identified for the second cohort using either the VT maps or the SUV images. In the third cohort, cerebellum and brainstem were identified as pseudo-reference regions, alongside subcortical white matter and temporal cortex. In addition, the SVM-based approach demonstrated robust performance even with a reduced number of subjects, therefore confirming its applicability in identifying pseudo-reference regions without a priori assumptions and with only limited data across different PET tracers.

Superpixel-ComBat modeling: A joint approach for harmonization and characterization of inter-scanner variability in T1-weighted images

Abstract T1-weighted imaging holds wide applications in clinical and research settings; however, the challenge of inter-scanner variability arises when combining data across scanners, which impedes multi-site research. To address this, post-acquisition harmonization methods such as statistical or deep learning approaches have been proposed to unify cross-scanner images. Nevertheless, how inter-scanner variability manifests in images and derived measures, and how to harmonize it in an interpretable manner, remains underexplored. To broaden our knowledge of inter-scanner variability and leverage it to develop a new harmonization strategy, we devised a pipeline to assess the interpretable inter-scanner variability in matched T1-weighted images across four 3T MRI scanners. The pipeline incorporates ComBat modeling with 3D superpixel parcellation algorithm (namely SP-ComBat), which estimates location and scale effects to quantify the shift and spread in relative signal distributions, respectively, concerning brain tissues in the image domain. The estimated parametric maps revealed significant contrast deviations compared to the joint signal distribution across scanners (p < 0.001), and the identified deviations in signal intensities may relate to differences in the inversion time acquisition parameter. To reduce the inter-scanner variability, we implemented a harmonization strategy involving proper image preprocessing and site effect removal by ComBat-derived parameters, achieving substantial improvement in image quality and significant reduction in variation of volumetric measures of brain tissues (p < 0.001). We also applied SP-ComBat to evaluate and characterize the performance of various image harmonization techniques, demonstrating a new way to assess image harmonization. In addition, we reported various metrics of T1-weighted images to quantify the impact of inter-scanner variation, including signal-to-noise ratio, contrast-to-noise ratio, signal inhomogeneity index, and structural similarity index. This study demonstrates a pipeline that extends the implementation of statistical ComBat method to the image domain in a practical manner for characterizing and harmonizing the inter-scanner variability in T1-weighted images, providing further insight for the studies focusing on the development of image harmonization methodologies and their applications.

Comparison of the variability and diagnostic efficacy of respiratory-gated PET/CT based radiomics features with ungated PET/CT in lung lesions

ObjectivesTo investigate the variability and diagnostic efficacy of respiratory-gated (RG) PET/CT based radiomics features compared to ungated (UG) PET/CT in the differentiation of non-small cell lung cancer (NSCLC) and benign lesions. Methods117 patients with suspected lung lesions from March 2020 to May 2021 and consent to undergo UG PET/CT and chest RG PET/CT (including phase-based quiescent period gating, pQPG and phase-matched 4D PET/CT, 4DRG) were prospectively included. 377 radiomics features were extracted from PET images of each scan. Paired t test was used to compare UG and RG features for inter-scan variability analysis. We developed three radiomics models with UG and RG features (i.e. UGModel, pQPGModel and 4DRGModel). ROC curves were used to compare diagnostic efficiencies, and the model-level comparison of diagnostic value was performed by five-fold cross-validation. A P value < 0.05 was considered as statistically significant. ResultsA total of 111 patients (average age ± standard deviation was 59.1 ± 11.6 y, range, 29 – 88 y, and 63 were males) with 209 lung lesions were analyzed for features variability and the subgroup of 126 non-metastasis lesions in 91 patients without treatment before PET/CT were included for diagnosis analysis. 101/377 (26.8 %) 4DRG features and 82/377 (21.8 %) pQPG features showed significant difference compared to UG features (both P<0.05). 61/377 (16.2 %) and 59/377 (15.6 %) of them showed significantly better discriminant ability (ΔAUC% (i.e. (AUCRG – AUCUG) / AUCUG×100 %) > 0 and P<0.05) in malignant recognition, respectively. For the model-level comparison, 4DRGModel achieved the highest diagnostic efficacy (sen 73.2 %, spe 87.3 %) compared with UGModel (sen 57.7 %, spe 76.4 %) and pQPGModel (sen 63.4 %, spe 81.8 %). ConclusionRG PET/CT performs better in the quantitative assessment of metabolic heterogeneity for lung lesions and the subsequent diagnosis in patients with NSCLC compared with UG PET/CT.

A deep learning model for brain segmentation across pediatric and adult populations.

Automated quantification of brain tissues on MR images has greatly contributed to the diagnosis and follow-up of neurological pathologies across various life stages. However, existing solutions are specifically designed for certain age ranges, limiting their applicability in monitoring brain development from infancy to late adulthood. This retrospective study aims to develop and validate a brain segmentation model across pediatric and adult populations. First, we trained a deep learning model to segment tissues and brain structures using T1-weighted MR images from 390 patients (age range: 2-81 years) across four different datasets. Subsequently, the model was validated on a cohort of 280 patients from six distinct test datasets (age range: 4-90 years). In the initial experiment, the proposed deep learning-based pipeline, icobrain-dl, demonstrated segmentation accuracy comparable to both pediatric and adult-specific models across diverse age groups. Subsequently, we evaluated intra- and inter-scanner variability in measurements of various tissues and structures in both pediatric and adult populations computed by icobrain-dl. Results demonstrated significantly higher reproducibility compared to similar brain quantification tools, including childmetrix, FastSurfer, and the medical device icobrain v5.9 (p-value< 0.01). Finally, we explored the potential clinical applications of icobrain-dl measurements in diagnosing pediatric patients with Cerebral Visual Impairment and adult patients with Alzheimer's Disease.

Robust thalamic nuclei segmentation from T1-weighted MRI using polynomial intensity transformation.

Accurate segmentation of thalamic nuclei, crucial for understanding their role in healthy cognition and in pathologies, is challenging to achieve on standard T1-weighted (T1w) magnetic resonance imaging (MRI) due to poor image contrast. White-matter-nulled (WMn) MRI sequences improve intrathalamic contrast but are not part of clinical protocols or extant databases. In this study, we introduce histogram-based polynomial synthesis (HIPS), a fast preprocessing transform step that synthesizes WMn-like image contrast from standard T1w MRI using a polynomial approximation for intensity transformation. HIPS was incorporated into THalamus Optimized Multi-Atlas Segmentation (THOMAS) pipeline, a method developed and optimized for WMn MRI. HIPS-THOMAS was compared to a convolutional neural network (CNN)-based segmentation method and THOMAS modified for the use of T1w images (T1w-THOMAS). The robustness and accuracy of the three methods were tested across different image contrasts (MPRAGE, SPGR, and MP2RAGE), scanner manufacturers (PHILIPS, GE, and Siemens), and field strengths (3T and 7T). HIPS-transformed images improved intra-thalamic contrast and thalamic boundaries, and HIPS-THOMAS yielded significantly higher mean Dice coefficients and reduced volume errors compared to both the CNN method and T1w-THOMAS. Finally, all three methods were compared using the frequently travelling human phantom MRI dataset for inter- and intra-scanner variability, with HIPS displaying the least inter-scanner variability and performing comparably with T1w-THOMAS for intra-scanner variability. In conclusion, our findings highlight the efficacy and robustness of HIPS in enhancing thalamic nuclei segmentation from standard T1w MRI.

Reduced cross-scanner variability using vendor-agnostic sequences for single-shell diffusion MRI.

To reduce the inter-scanner variability of diffusion MRI (dMRI) measures between scanners from different vendors by developing a vendor-neutral dMRI pulse sequence using the open-source vendor-agnostic Pulseq platform. We implemented a standard EPI based dMRI sequence in Pulseq. We tested it on two clinical scanners from different vendors (Siemens Prisma and GE Premier), systematically evaluating and comparing the within- and inter-scanner variability across the vendors, using both the vendor-provided and Pulseq dMRI sequences. Assessments covered both a diffusion phantom and three human subjects, using standard error (SE) and Lin's concordance correlation to measure the repeatability and reproducibility of standard DTI metrics including fractional anisotropy (FA) and mean diffusivity (MD). Identical dMRI sequences were executed on both scanners using Pulseq. On the phantom, the Pulseq sequence showed more than a 2.5× reduction in SE (variability) across Siemens and GE scanners. Furthermore, Pulseq sequences exhibited markedly reduced SE in-vivo, maintaining scan-rescan repeatability while delivering lower variability in FA and MD (more than 50% reduction in cortical/subcortical regions) compared to vendor-provided sequences. The Pulseq diffusion sequence reduces the cross-scanner variability for both phantom and in-vivo data, which will benefit multi-center neuroimaging studies and improve the reproducibility of neuroimaging studies.

Predicting long-term progression of Alzheimer’s disease using a multimodal deep learning model incorporating interaction effects

BackgroundIdentifying individuals with mild cognitive impairment (MCI) at risk of progressing to Alzheimer’s disease (AD) provides a unique opportunity for early interventions. Therefore, accurate and long-term prediction of the conversion from MCI to AD is desired but, to date, remains challenging. Here, we developed an interpretable deep learning model featuring a novel design that incorporates interaction effects and multimodality to improve the prediction accuracy and horizon for MCI-to-AD progression.MethodsThis multi-center, multi-cohort retrospective study collected structural magnetic resonance imaging (sMRI), clinical assessments, and genetic polymorphism data of 252 patients with MCI at baseline from the Alzheimer’s Disease Neuroimaging Initiative (ADNI) database. Our deep learning model was cross-validated on the ADNI-1 and ADNI-2/GO cohorts and further generalized in the ongoing ADNI-3 cohort. We evaluated the model performance using the area under the receiver operating characteristic curve (AUC), accuracy, sensitivity, specificity, and F1 score.ResultsOn the cross-validation set, our model achieved superior results for predicting MCI conversion within 4 years (AUC, 0.962; accuracy, 92.92%; sensitivity, 88.89%; specificity, 95.33%) compared to all existing studies. In the independent test, our model exhibited consistent performance with an AUC of 0.939 and an accuracy of 92.86%. Integrating interaction effects and multimodal data into the model significantly increased prediction accuracy by 4.76% (P = 0.01) and 4.29% (P = 0.03), respectively. Furthermore, our model demonstrated robustness to inter-center and inter-scanner variability, while generating interpretable predictions by quantifying the contribution of multimodal biomarkers.ConclusionsThe proposed deep learning model presents a novel perspective by combining interaction effects and multimodality, leading to more accurate and longer-term predictions of AD progression, which promises to improve pre-dementia patient care.

A Comparative Study of Three Systems for Liver Magnetic Resonance Elastography.

Different MR elastography (MRE) systems may produce different stiffness measurements, making direct comparison difficult in multi-center investigations. To assess the repeatability and reproducibility of liver stiffness measured by three typical MRE systems. Prospective. Thirty volunteers without liver disease history (20 males, aged 21-28)/5 gel phantoms. 3.0 T United Imaging Healthcare (UIH), 1.5 T Siemens Healthcare, 3.0 T General Electric Healthcare (GE)/Echo planar imaging-based MRE sequence. Wave images of volunteers and phantoms were acquired by three MRE systems. Tissue stiffness was evaluated by two observers, while phantom stiffness was assessed automatically by code. The reproducibility across three MRE systems was quantified based on the mean stiffness of each volunteer and phantom. Intraclass correlation coefficients (ICC), coefficients of variation (CV), and Bland-Altman analyses were used to assess the interobserver reproducibility, the interscan repeatability, and the intersystem reproducibility. Paired t-tests were performed to assess the interobserver and interscan variation. Friedman tests with Dunn's multiple comparison correction were performed to assess the intersystem variation. P values less than 0.05 indicated significant difference. The reproducibility of stiffness measured by the two observers demonstrated consistency with ICC > 0.92, CV < 4.32%, Mean bias < 2.23%, and P > 0.06. The repeatability of measurements obtained using the electromagnetic system for the liver revealed ICC > 0.96, CV < 3.86%, Mean bias < 0.19%, P > 0.90. When considering the range of reproducibility across the three systems for liver evaluations, results ranged with ICCs from 0.70 to 0.87, CVs from 6.46% to 10.99%, and Mean biases between 1.89% and 6.30%. Phantom studies showed similar results. The values of measured stiffness differed across all three systems significantly. Liver stiffness values measured from different MRE systems can be different, but the measurements across the three MRE systems produced consistent results with excellent reproducibility. 1 TECHNICAL EFFICACY: Stage 2.

Assessment of Feasibility and Interscan Variability of Short-time Cardiac MRI for Cardiotoxicity Evaluation in Breast Cancer.

Purpose To investigate the feasibility and interscan variability of short-time cardiac MRI protocol after chemotherapy in individuals with breast cancer. Materials and Methods A total of 13 healthy female controls (mean age, 52.4 years ± 13.2 [SD]) and 85 female participants with breast cancer (mean age, 51.8 years ± 9.9) undergoing chemotherapy prospectively underwent routine breast MRI and short-time cardiac MRI using a 3-T scanner with peripheral pulse gating in the prone position. Interscan, intercoil, and interobserver reproducibility and variability of native T1 and extracellular volume (ECV), as well as ventricular functional parameters, were measured using the intraclass correlation coefficient (ICC), standard error of measurement (SEM), or coefficient of variation (CoV). Results Left ventricular functional parameters had excellent interscan reproducibility (ICC ≥ 0.80). Left ventricular ejection fraction showed low interscan variability in control and chemotherapy participants (SEM, 2.0 and 1.2; CoV, 3.1 and 1.9, respectively). Native T1 showed excellent interscan (ICC, 0.75) and intercoil (ICC, 0.81) reproducibility in the control group and good interscan reproducibility (ICC, 0.72 and 0.73, respectively) in the participants undergoing immediate and remote chemotherapy. Interscan reproducibility for ECV was excellent in the control group and in the remote chemotherapy group (ICC, 0.93 and 0.88, respectively) and fair in the immediate chemotherapy group (ICC, 0.52). In the regional analysis, interscan repeatability and variability of native T1 and ECV were superior in the anteroseptum or inferoseptum than in other segments in the immediate chemotherapy group. Native T1 and ECV had good to excellent interobserver agreement across all groups. Conclusion Short-time cardiac MRI showed excellent results for interscan, intercoil, and interobserver reproducibility and variability for ventricular functional or tissue characterization parameters, suggesting that this modality is feasible for routine surveillance of cardiotoxicity evaluation in individuals with breast cancer. Keywords: Cardiac MRI, Heart, Cardiomyopathy ClinicalTrials.gov registration no. NCT03301389 Supplemental material is available for this article. © RSNA, 2024.

Receive coil quality assurance procedure and automated analysis for ViewRay MRIdian MR-Linac.

Regular receiving coil quality assurance (QA) is required to ensure image quality of an MRIdian Linac system. The manufacturer provides a spherical phantom and positioning tube for single-slice signal-to-noise ratio (SNR) and uniformity assessments. We aimed to improve imaging setup and coverage and eliminate inter-scan variability by employing multi-slice imaging of a stable phantom. Additionally, we strived to expedite analysis by developing objective, automated analysis software. A 5300mL cylindrical plastic bottle placed in plastic bins was scanned at isocenter using a spin-echo sequence with NEMA-recommended parameters and 18 axial slices, avoiding phantom repositioning. Acquisition was repeated with and without prescan normalization filtering and by saving uncombined element images. Obtained data were analyzed using custom open-source MATLAB code. Signal and noise images were automatically assigned, and ROIs for SNR and uniformity calculations were defined using image thresholding. SNR and uniformity pass/fail decisions were made using baseline comparisons. The proposed method was successfully implemented as monthly coil QA for 3.5 years. Setup and scanning took 41 min on average for a coil set. Automated image analysis was completed in a few minutes. Signal intensity peaked around +90 or -90mm for Torso or Head/Neck coil unfiltered images. Noise peaked and minimized SNR inside±30mm from isocenter, while maximizing it around±130mm. Prescan normalization smoothed signal response, reduced SNR and increased uniformity. Individual coil element image analysis identified their position, signal or noise response and SNR. SNR and uniformity pass/fail thresholds were set for already tested and new coils. Conspicuous and subtle Torso coil malfunctions were detected considering baseline deviations of combined and individual element results. Our QA method eliminated observer bias and provided insights into coil function, image filtering performance and coil element location. It provided SNR and uniformity thresholds and identified faulty coil elements.

No change in network connectivity measurements between separate rsfMRI acquisition times.

The role of resting state functional MRI (rsfMRI) is increasing in the field of epilepsy surgery because it is possible to interpolate network connectivity patterns across the brain with a high degree of spatial resolution. Prior studies have shown that by rsfMRI with scalp electroencephalography (EEG), an epileptogenic network can be modeled and visualized with characteristic patterns of connectivity that are relevant to both seizure-related and neuropsychological outcomes after surgery. The aim of this study is to show that a 5-min acquisition time provides reproducible results related to the relevant connectivity metrics when compared to a separately acquired 5-min scan. Fourteen separate rsfMRI sessions from ten different patients were used for comparison, comprised of patients with temporal lobe epilepsy both pre- and post-operation. Results showed that there was no significant difference in any of the connectivity metrics when comparing both 5-min scans to each other. These data support the continued use of a 5-min scan for epileptogenic network modeling in future studies because the inter-scan variability is sufficiently low as not to alter the output metrics characterizing the network connectivity.

Inter-scanner comparability of Z-scores for native myocardial T1 and T2 mapping

BackgroundCardiovascular Magnetic Resonance (CMR) native T1 and T2 mapping serve as robust, contrast-agent-free diagnostic tools, but hardware- and software-specific sources of variability limit the generalizability of data across CMR platforms, consequently limiting the interpretability of patient-specific parametric data. Z-scores are used to describe the relationship of observed values to the mean results as obtained in a sufficiently large normal sample. They have been successfully used to describe the severity of quantifiable abnormalities in medicine, specifically in children and adolescents. The objective of this study was to observe whether z-scores can improve the comparability of T1 and T2 mapping values across CMR scanners, field strengths, and sequences from different vendors in the same participant rather than different participants (as seen in previous studies). MethodsFifty-one healthy volunteers (26 men/25 women, mean age = 43 ± 13.51) underwent three CMR exams on three different scanners, using a Modified Look-Locker Inversion Recovery (MOLLI) 5-(3)− 3 sequence to quantify myocardial T1. For T2 mapping, a True Fast Imaging with steady-state free precession (TRUFI) sequence was used on a 3 T Skyra™ (Siemens), and a T2 Fast Spin Echo (FSE) sequence was used on 1.5 T Artist™ (GE) and 3.0 T Premier™ (GE) scanners. The averages of basal and mid-ventricular short axis slices were used to derive means and standard deviations of global mapping values. We used intra-class comparisons (ICC), repeated measures ANOVA, and paired Student’s t-tests for statistical analyses. ResultsThere was a significant improvement in intra-subject comparability of T1 (ICC of 0.11 (95% CI= −0.018, −0.332) vs 0.78 (95% CI= 0.650, 0.866)) and T2 (ICC of 0.35 (95% CI= −0.053, 0.652) vs 0.83 (95% CI= 0.726, 0.898)) when using z-scores across all three scanners. While the absolute global T1 and T2 values showed a statistically significant difference between scanners (p < 0.001), no such differences were identified using z-scores (T1z: p = 0.771; T2z: p = 0.985). Furthermore, when images were not corrected for motion, T1 z-scores showed significant inter-scanner variability (p < 0.001), resolved by motion correction. ConclusionEmploying z-scores for reporting myocardial T1 and T2 removes the variation of quantitative mapping results across different MRI systems and field strengths, improving the clinical utility of myocardial tissue characterization in patients with suspected myocardial disease.

Robustness of apparent diffusion coefficient–based lymph node classification for diagnosis of prostate cancer metastasis

ObjectivesThe aim of this proof-of-principle study combining data analysis and computer simulation was to evaluate the robustness of apparent diffusion coefficient (ADC) values for lymph node classification in prostate cancer under conditions comparable to clinical practice.Materials and methodsTo assess differences in ADC and inter-rater variability, ADC values of 359 lymph nodes in 101 patients undergoing simultaneous prostate-specific membrane antigen (PSMA)-PET/MRI were retrospectively measured by two blinded readers and compared in a node-by-node analysis with respect to lymph node status. In addition, a phantom and 13 patients with 86 lymph nodes were prospectively measured on two different MRI scanners to analyze inter-scanner agreement. To estimate the diagnostic quality of the ADC in real-world application, a computer simulation was used to emulate the blurring caused by scanner and reader variability. To account for intra-individual correlation, the statistical analyses and simulations were based on linear mixed models.ResultsThe mean ADC of lymph nodes showing PSMA signals in PET was markedly lower (0.77 × 10−3 mm2/s) compared to inconspicuous nodes (1.46 × 10−3 mm2/s, p < 0.001). High inter-reader agreement was observed for ADC measurements (ICC 0.93, 95%CI [0.92, 0.95]). Good inter-scanner agreement was observed in the phantom study and confirmed in vivo (ICC 0.89, 95%CI [0.84, 0.93]). With a median AUC of 0.95 (95%CI [0.92, 0.97]), the simulation study confirmed the diagnostic potential of ADC for lymph node classification in prostate cancer.ConclusionOur model-based simulation approach implicates a high potential of ADC for lymph node classification in prostate cancer, even when inter-rater and inter-scanner variability are considered.Clinical relevance statementThe ADC value shows a high diagnostic potential for lymph node classification in prostate cancer. The robustness to scanner and reader variability implicates that this easy to measure and widely available method could be readily integrated into clinical routine.Key Points• The diagnostic value of the apparent diffusion coefficient (ADC) for lymph node classification in prostate cancer is unclear in the light of inter-rater and inter-scanner variability.• Metastatic and inconspicuous lymph nodes differ significantly in ADC, resulting in a high diagnostic potential that is robust to inter-scanner and inter-rater variability.• ADC has a high potential for lymph node classification in prostate cancer that is maintained under conditions comparable to clinical practice.Graphical abstract

Pulsed wave Doppler ultrasound: Accuracy, variability, and impact of acquisition parameters on flow measurements.

Pulsed wave Doppler ultrasound is a useful modality for assessing vascular health as it quantifies blood flow characteristics. To facilitate accurate diagnosis, accuracy and consistency of this modality should be assessed through Doppler quality assurance (QA). The purpose of this study was to characterize the accuracy, reproducibility, and inter-scanner variability of ultrasound flow velocity measurements via a flow phantom, with a focus on the effect of systematic acquisition parameters on measured flow velocity accuracy. Using a manufacturer-calibrated flow phantom, pulsed wave measurements were acquired on five clinical systems (iU22, Philips) with three models of transducers, including both linear and curvilinear models. The peak and mean flow velocities were estimated by vendor-supplied spectral analysis tools. To investigate intra- and inter-scanner variability, measurements were repeated using each scanner-transducer pair under a standardized set of conditions. Inter-scanner variability was assessed using ANOVA. Flow velocity accuracy was investigated by mean absolute percentage error. The impacts of receive gain, measurement depth, and beam steering on measured flow velocity accuracy were examined by varying each parameter over its available range and comparing to the ground truth flow velocity. Inter-scanner variability was statistically significant for peak flow measurements made using both linear and curvilinear transducers, though absolute differences in measured velocity were small. Inter-scanner variability was not statistically significant for mean flow velocity. Receive gain, measurement depth, and beam steering were all found to impact the accuracy of measured flow characteristics for linear transducers. Accuracy of the flow measurements made with the curvilinear transducer demonstrated high consistency to changes in receive gain at a constant depth, though were impacted by increasing the measurement depth. Carefully and consistently selected acquisition and set-up parameters are essential in order to establish a reliable and meaningful QA program.

Assessment of resting myocardial blood flow in regions of known transmural scar to confirm accuracy and precision of 3D cardiac positron emission tomography

BackgroundComposite invasive and non-invasive data consistently demonstrate that resting myocardial blood flow (rMBF) in regions of known transmural myocardial scar (TMS) converge on a value of ~ 0.30 mL/min/g or lower. This value has been confirmed using the 3 most common myocardial perfusion agents (13N, 15O-H2O and 82Rb) incorporating various kinetic models on older 2D positron emission tomography (PET) systems. Thus, rMBF in regions of TMS can serve as a reference “truth” to evaluate low-end accuracy of various PET systems and software packages (SWPs). Using 82Rb on a contemporary 3D-PET-CT system, we sought to determine whether currently available SWP can accurately and precisely measure rMBF in regions of known TMS.ResultsMedian rMBF (in mL/min/g) and COV in regions of TMS were 0.71 [IQR 0.52–1.02] and 0.16 with 4DM; 0.41 [0.34–0.54] and 0.10 with 4DM-FVD; 0.66 [0.51–0.85] and 0.11 with Cedars; 0.51 [0.43–0.61] and 0.08 with Emory-Votaw; 0.37 [0.30–0.42], 0.07 with Emory-Ottawa, and 0.26 [0.23–0.32], COV 0.07 with HeartSee.ConclusionsSWPs varied widely in low end accuracy based on measurement of rMBF in regions of known TMS. 3D PET using 82Rb and HeartSee software accurately (0.26 mL/min/g, consistent with established values) and precisely (COV = 0.07) quantified rMBF in regions of TMS. The Emory-Ottawa software yielded the next-best accuracy (0.37 mL/min/g), though rMBF was higher than established gold-standard values in ~ 5% of the resting scans. 4DM, 4DM-FDV, Cedars and Emory-Votaw SWP consistently resulted values higher than the established gold standard (0.71, 0.41, 0.66, 0.51 mL/min/g, respectively), with higher interscan variability (0.16, 0.11, 0.11, and 0.09, respectively).Trial registration: clinicaltrial.gov, NCT05286593, Registered December 28, 2021, https://clinicaltrials.gov/ct2/show/NCT05286593.

Dual-energy index variation when evaluating the potential ferromagnetism of ex vivo bullets.

An MRI is potentially hazardous for patients with retained ferromagnetic bullets. Recent studies have aimed to develop dual-energy computed tomography (DECT) as a screening tool for recognising highly ferromagnetic bullets. Inconsistent findings have been ascribed to inherent CT technology differences. Previous research demonstrated significant Hounsfield unit (HU) measurement variation among single-source CT machines. This study investigated the theoretical dual-energy index (DEI) variation between DECT machines when evaluating the potential ferromagnetic properties within the same sample of ex vivo bullets and metal phantoms. An experimental ex vivo study was conducted on eight metal phantoms and 10 unused bullets individually positioned in the same Perspex head phantom and scanned on two DECT machines. Two senior radiology registrars independently recorded the HU readings, and DEI values were calculated. Statistical analysis was performed using non-parametric methods for paired data, namely the Signed Rank Test. The DEI values based on mean HU readings between the DECT machines were compared. Inter- and intra-reader agreement was not statistically significant. The metal phantoms had poor interscanner agreement, with an overlap of the ferromagnetic and non-ferromagnetic ranges. The bullets had good interscanner agreement, with a similar ferromagnetic to non-ferromagnetic relationship. The use of DEI values negates the previous assumption that significant interscanner variability exists among different DECT technologies while assessing highly attenuative ex vivo bullets. This investigation demonstrated that even though HU readings may be variable, the implementation of the DEI equation translates this into comparable values with good interscanner agreement.

Inter-software and inter-scan variability in measurement of epicardial adipose tissue: a three-way comparison of a research-specific, a freeware and a coronary application software platform

ObjectivesEpicardial adipose tissue (EAT) is a proposed marker of cardiovascular risk; however, clinical application may be limited by variability in post-processing software platforms. We assessed inter-vendor agreement of EAT volume (EATv) and attenuation on both contrast-enhanced (CE) and non-contrast CT (NCT) using a standard coronary CT reporting software (Vitrea), an EAT research-specific software (QFAT) and a freeware imaging software (OsiriX).MethodsSeventy-six consecutive patients undergoing simultaneous CE and NCT had complete volumetric EAT measurement. Between-software, within-software NCT vs. CE, and inter- and intra-observer agreement were evaluated with analysis by ANOVA (with post hoc adjustment), Bland-Altman with 95% levels of agreement (LoA) and intraclass correlation coefficient (ICC).ResultsMean EATv (freeware 53 ± 31 mL vs. research 93 ± 43 mL vs. coronary 157 ± 64 mL) and attenuation (freeware − 72 ± 25 HU vs. research − 75 ± 3 HU vs. coronary − 61 ± 10 HU) were significantly different between all vendors (ANOVA p < 0.001). EATv was consistently higher in NCT vs. CE for all software packages, with most reproducibility found in research software (bias 26 mL, 95% LoA: 2 to 56 mL), compared to freeware (bias 11 mL 95% LoA: − 46 mL to 69 mL) and coronary software (bias 10 mL 95% LoA: − 127 to 147 mL). Research software had more comparable NCT vs. CE attenuation (− 75 vs. − 72 HU) compared to freeware (− 72 vs. − 57 HU) and coronary (− 61 vs. − 39 HU). Excellent inter-observer agreement was seen with research (ICC 0.98) compared to freeware (ICC 0.73) and coronary software (ICC 0.75) with narrow LoA on Bland-Altman analysis.ConclusionThere are significant inter-vendor differences in EAT assessment. Our study suggests that research-specific software has better agreement and reproducibility compared to freeware or coronary software platforms.Key Points• There are significant differences between EAT volume and attenuation values between software platforms, regardless of scan type.• Non-contrast scans routinely have higher mean EAT volume and attenuation; however, this finding is only consistently seen with research-specific software.• Of the three analyzed packages, research-specific software demonstrates the highest reproducibility, agreement, and reliability for both inter-scan and inter-observer agreement.

Same-day repeatability and Between-Sequence reproducibility of Mean ADC in PI-RADS lesions

PurposeThis study aimed to assess repeatability after repositioning (inter-scan), intra-rater, inter-rater and inter-sequence variability of mean apparent diffusion coefficient (ADC) measurements in MRI-detected prostate lesions. MethodForty-three patients with suspicion for prostate cancer were included and received a clinical prostate bi-/multiparametric MRI examination with repeat scans of the T2-weighted and two DWI-weighted sequences (ssEPI and rsEPI). Two raters (R1 and R2) performed single-slice, 2D regions of interest (2D-ROIs) and 3D-segmentation-ROIs (3D-ROIs). Mean bias, corresponding limits of agreement (LoA), mean absolute difference, within-subject coefficient of variation (CoV) and repeatability/reproducibility coefficient (RC/RDC) were calculated. Bradley & Blackwood test was used for variance comparison. Linear mixed models (LMM) were used to account for multiple lesions per patient. ResultsInter-scan repeatability, intra-rater and inter-sequence reproducibility analysis of ADC showed no significant bias. 3D-ROIs demonstrated significantly less variability than 2D-ROIs (p < 0.01). Inter-rater comparison demonstrated small significant systematic bias of 57 × 10-6 mm2/s for 3D-ROIs (p < 0.001). Intra-rater RC, with the lowest variation, was 145 and 189 × 10-6 mm2/s for 3D- and 2D-ROIs, respectively. For 3D-ROIs of ssEPI, RCs and RDCs were 190–198 × 10-6 mm2/s for inter-scan, inter-rater and inter-sequence variation. No significant differences were found for inter-scan, inter-rater and inter-sequence variability. ConclusionsIn a single-scanner setting, single-slice ADC measurements showed considerable variation, which may be lowered using 3D-ROIs. For 3D-ROIs, we propose a cut-off of ∼ 200 × 10-6 mm2/s for differences introduced by repositioning, rater or sequence effects. The results suggest that follow-up measurements should be possible by different raters or sequences.

Limited impact of discretization/interpolation parameters on the predictive power of CT radiomic features in a surgical cohort of pancreatic cancer patients.

To explore the variation of the discriminative power of CT (Computed Tomography) radiomic features (RF) against image discretization/interpolation in predicting early distant relapses (EDR) after upfront surgery. Data of 144 patients with pre-surgical high contrast CT were processed consistently with IBSI (Image Biomarker Standardization Initiative) guidelines. Image interpolation/discretization parameters were intentionally changed, including cubic voxel size (0.21-27 mm3) and binning (32-128 grey levels) in a 15 parameter's sets. After excluding RF with poor inter-observer delineation agreement (ICC < 0.80) and not negligible inter-scanner variability, the variation of 80 RF against discretization/interpolation was first quantified. Then, their ability in classifying patients with early distant relapses (EDR, < 10months, previously assessed at the first quartile value of time-to-relapse) was investigated in terms of AUC (Area Under Curve) variation for those RF significantly associated to EDR. Despite RF variability against discretization/interpolation parameters was large and only 30/80 RF showed %COV < 20 (%COV = 100*STDEV/MEAN), AUC changes were relatively limited: for 30 RF significantly associated with EDR (AUC values around 0.60-0.70), the mean values of SD of AUC variability and AUC range were 0.02 and 0.05 respectively. AUC ranges were between 0.00 and 0.11, with values ≤ 0.05 in 16/30 RF. These variations were further reduced when excluding the extreme values of 32 and 128 for grey levels (Average AUC range 0.04, with values between 0.00 and 0.08). The discriminative power of CT RF in the prediction of EDR after upfront surgery for pancreatic cancer is relatively invariant against image interpolation/discretization within a large range of voxel sizes and binning.