Fully-automated quality assurance in multi-center studies using MRI phantom measurements

Deep Learning Approaches for Automatic Quality Assurance of Magnetic Resonance Images Using ACR Phantom

BACKGROUND

In recent years, there has been a growing trend towards utilizing Artificial Intelligence (AI) and machine learning techniques in medical imaging, including for the purpose of automating quality assurance. In this research, we aimed to develop and evaluate various deep learning-based approaches for automatic quality assurance of Magnetic Resonance (MR) images using the American College of Radiology (ACR) standards.

METHODS

The study involved the development, optimization, and testing of custom convolutional neural network (CNN) models. Additionally, popular pre-trained models such as VGG16, VGG19, ResNet50, InceptionV3, EfficientNetB0, and EfficientNetB5 were trained and tested. The use of pre-trained models, particularly those trained on the ImageNet dataset, for transfer learning was also explored. Two-class classification models were employed for assessing spatial resolution and geometric distortion, while an approach classifying the image into 10 classes representing the number of visible spokes was used for the low contrast.

RESULTS

Our results showed that deep learning-based methods can be effectively used for MR image quality assurance and can improve the performance of these models. The low contrast test was one of the most challenging tests within the ACR phantom.

CONCLUSIONS

Overall, for geometric distortion and spatial resolution, all of the deep learning models tested produced prediction accuracy of 80% or higher. The study also revealed that training the models from scratch performed slightly better compared to transfer learning. For the low contrast, our investigation emphasized the adaptability and potential of deep learning models. The custom CNN models excelled in predicting the number of visible spokes, achieving commendable accuracy, recall, precision, and F1 scores.

Tubers Affecting the Fusiform Face Area Are Associated with Autism Diagnosis

OBJECTIVE

Tuberous sclerosis complex (TSC) is associated with focal brain "tubers" and a high incidence of autism spectrum disorder (ASD). The location of brain tubers associated with autism may provide insight into the neuroanatomical substrate of ASD symptoms.

METHODS

We delineated tuber locations for 115 TSC participants with ASD (n = 31) and without ASD (n = 84) from the Tuberous Sclerosis Complex Autism Center of Excellence Research Network. We tested for associations between ASD diagnosis and tuber burden within the whole brain, specific lobes, and at 8 regions of interest derived from the ASD neuroimaging literature, including the anterior cingulate, orbitofrontal and posterior parietal cortices, inferior frontal and fusiform gyri, superior temporal sulcus, amygdala, and supplemental motor area. Next, we performed an unbiased data-driven voxelwise lesion symptom mapping (VLSM) analysis. Finally, we calculated the risk of ASD associated with positive findings from the above analyses.

RESULTS

There were no significant ASD-related differences in tuber burden across the whole brain, within specific lobes, or within a priori regions derived from the ASD literature. However, using VLSM analysis, we found that tubers involving the right fusiform face area (FFA) were associated with a 3.7-fold increased risk of developing ASD.

INTERPRETATION

Although TSC is a rare cause of ASD, there is a strong association between tuber involvement of the right FFA and ASD diagnosis. This highlights a potentially causative mechanism for developing autism in TSC that may guide research into ASD symptoms more generally. ANN NEUROL 2023;93:577-590.

Machine learning framework for automatic image quality evaluation involving a mammographic American College of Radiology phantom

PURPOSE

The image quality (IQ) of mammographic images is essential when making a diagnosis, but the quality assurance process for radiological equipment is subjective. We therefore aimed to design an automatic IQ evaluation architecture based on a support vector machine (SVM) dedicated to evaluating images taken of mammography American College of Radiology (ACR) phantom.

METHODS

A total of 461 phantom images were acquired using mammographic equipment from 10 vendors. Two experienced medical physicists scored the images by consensus. The phantom datasets were randomly divided into training (80%) and testing (20%) sets. Each phantom image (with 6 fibers, 5 specks, and 5 masses) was detected by using bounding boxes, then cropped and divided into 16 pattern images. We identified 159 features for each pattern image. Manual scores were used to assign 3 labels (visible, invisible, and semivisible) to each pattern image. Multiclass-SVM models were trained with 3 types of patterns. Sub-datasets were randomly selected at 10% increments of the total dataset to determine a minimal effective training subset size for the automatic framework. A feature combination test and an analysis of variance were performed to identify the most influential features.

RESULTS

The accuracy of the model in evaluating fiber, speck, and mass patterns was 90.2%, 98.2%, and 88.9%, respectively. The performance was equivalent when the sample size was at least 138 (30% of 461) phantom images. The most influential feature was the position feature.

CONCLUSIONS

The proposed SVM-based automatic IQ evaluation framework applied to a mammographic ACR phantom accurately matched manual evaluations.

Orbitofrontal control of conduct problems? Evidence from healthy adolescents processing negative facial affect

Conduct problems (CP) in patients with disruptive behavior disorders have been linked to impaired prefrontal processing of negative facial affect compared to controls. However, it is unknown whether associations with prefrontal activity during affective face processing hold along the CP dimension in a healthy population sample, and how subcortical processing is affected. We measured functional brain responses during negative affective face processing in 1444 healthy adolescents [M = 14.39 years (SD = 0.40), 51.5% female] from the European IMAGEN multicenter study. To determine the effects of CP, we applied a two-step approach: (a) testing matched subgroups of low versus high CP, extending into the clinical range [N = 182 per group, M = 14.44 years, (SD = 0.41), 47.3% female] using analysis of variance, and (b) considering (non)linear effects along the CP dimension in the full sample and in the high CP group using multiple regression. We observed no significant cortical or subcortical effect of CP group on brain responses to negative facial affect. In the full sample, regression analyses revealed a significant linear increase of left orbitofrontal cortex (OFC) activity with increasing CP up to the clinical range. In the high CP group, a significant inverted u-shaped effect indicated that left OFC responses decreased again in individuals with high CP. Left OFC activity during negative affective processing which is increasing with CP and decreasing in the highest CP range may reflect on the importance of frontal control mechanisms that counteract the consequences of severe CP by facilitating higher social engagement and better evaluation of social content in adolescents.

Image quality assessments according to the angle of tilt of a flex tilt coil supporting device: An ACR phantom study

In this study, we assessed how image quality depends on the angle of tilt of a flex tilt coil supporting device during an MRI examination. All measurements were performed with an American College of Radiology (ACR) MRI phantom using a flex tilt coil supporting device. All images were analyzed using an automatic assessment method following the ACR MRI accreditation guidance. Image quality was compared between acquisitions grouped according to the angle of tilt of the coil supporting device: group A (Flat mode), group B (10˚), and group C (18˚). All measured image qualities were within the ACR recommended criteria, regardless of the angle of tilt of the flex tilt coil supporting device. However, statistically significant differences between the three groups were found for slice thickness, position accuracy, image intensity uniformity, and SNR (P < 0.05, ANOVA). The flex tilt coil supporting device can provide sufficient image quality, passing the criteria of the ACR MRI guideline, despite differences in slice thickness, slice position accuracy, image intensity uniformity, and SNR according to the angle of tilt.

Automatic MR image quality evaluation using a Deep CNN: A reference-free method to rate motion artifacts in neuroimaging

Motion artifacts on magnetic resonance (MR) images degrade image quality and thus negatively affect clinical and research scanning. Considering the difficulty in preventing patient motion during MR examinations, the identification of motion artifact has attracted significant attention from researchers. We propose an automatic method for the evaluation of motion corrupted images using a deep convolutional neural network (CNN). Deep CNNs has been used widely in image classification tasks. While such methods require a significant amount of annotated training data, a scarce resource in medical imaging, the transfer learning and fine-tuning approaches allow us to use a smaller amount of data. Here we selected four renowned architectures, initially trained on Imagenet contest dataset, to fine-tune. The models were fine-tuned using patches from an annotated dataset composed of 68 T1-weighted volumetric acquisitions from healthy volunteers. For training and validation 48 images were used, while the remaining 20 images were used for testing. Each architecture was fine-tuned for each MR axis, detecting the motion artifact per patches from the three orthogonal MR acquisition axes. The overall average accuracy for the twelve models (three axes for each of four architecture) was 86.3%. As our goal was to detect fine-grained corruption in the image, we performed an extensive search on lower layers from each of the four architectures, since they filter small regions in the original input. Experiments showed that architectures with fewer layers than the original ones reported the better results for image patches with an overall average accuracy of 90.4%. The accuracies per architecture were similar so we decided to explore all four architectures performing a result consensus. Also, to determine the probability of motion artifacts presence on the whole acquisition a combination of the three axes were performed. The final architecture consists of an artificial neural network (ANN) classifier combining all models from the four shallower architectures, which overall acquisition-based accuracy was 100.0%. The proposed method generalization was tested using three different MR data: (1) MR image acquired in epilepsy patients (93 acquisitions); (2) MR image presenting susceptibility artifact (22 acquisitions); and (3) MR image acquired from different scanner vendor (20 acquisitions). The achieved acquisition-based accuracy on generalization tests (1) 90.3%, (2) 63.6%, and (3) 75.0%) suggests that domain adaptation is necessary. Our proposed method can be rapidly applied to large amounts of image data, providing a motion probability p∈[0,1] per acquisition. This method output can be used as a scale to identify the motion corrupted images from the dataset, thus minimizing the time spent on visual quality control.

Technical Note: Validation of an automatic ACR phantom quality assurance tool for an MR-guided radiotherapy system

PURPOSE

To modify and evaluate an automatic American College of Radiology (ACR) phantom analysis toolbox for ACR quality assurance (QA) on a low-field MR-guided radiotherapy system (ViewRay).

METHODS

An open-source toolbox was modified for ACR QA of a 0.35T MRI system (ViewRay MRIdian). A total of 17 ACR datasets were evaluated, including 10 datasets acquired from different systems across the world, and seven datasets acquired at our center between 2014 and 2020. All required ACR tests, geometric accuracy (GA), high-contrast spatial resolution (HCSR), slice thickness accuracy (ST), slice position accuracy (SP), percent integral uniformity (PIU), percentage signal ghosting (PSG), and low-contrast object detectability (LCOD), were assessed manually and using the toolbox automatically. Measurements between manual and automatic analysis were compared. Precision, recall, and accuracy were calculated, where the manual results were used as the ground truth.

RESULTS

The software took less than 2 min to complete all seven tests, which usually requires 40 min or more if analyzed manually. Overall, the automatic measurement was consistent with the manual result. The absolute differences between the two measurements were 0.72 ± 0.66 (mm), 0.01 ± 0.03, 0.50 ± 0.60 (mm), 0.39 ± 0.41 (mm), 1.01 ± 1.00 (%), 0.0016 ± 0.0019, and 2.79 ± 2.29 for the seven tests. The precision of the automatic toolbox was 100% for all tests, indicating that a test would 100% pass a manual analysis if it had passed the automatic analysis. Recall and accuracy were ≥ 96% for GA, HCSR, SP, PIU, PSG, and LCOD tests, and 91% for the ST test.

CONCLUSION

An automatic ACR QA tool was adopted and evaluated for the low-field MR-guided radiotherapy (MRgRT) system. Overall, the toolbox provided comparable results as manual analysis, and reduced the processing time from over 40 min to <2 min. This toolbox holds the potential to be widely adopted either as a second check tool or partially replace human measurement for MRgRT programs using the same system.

Classifying MRI motion severity using a stacked ensemble approach

Motion artifacts are a common occurrence in Magnetic Resonance Imaging exam. Motion during acquisition has a profound impact on workflow efficiency, often requiring a repeat of sequences. Furthermore, motion artifacts may escape notice by technologists, only to be revealed at the time of reading by the radiologists, affecting their diagnostic quality. There is a paucity of clinical tools to identify and quantitatively assess the severity of motion artifacts in MRI. An image with subtle motion may still have diagnostic value, while severe motion may be uninterpretable by radiologists and requires the exam to be repeated. Therefore, a tool for the automatic identification of motion artifacts would aid in maintaining diagnostic quality, while potentially driving workflow efficiencies. Here we aim to quantify the severity of motion artifacts from MRI images using deep learning. Impact of subject movement parameters like displacement and rotation on image quality is also studied. A state-of-the-art, stacked ensemble model was developed to classify motion artifacts into five levels (no motion, slight, mild, moderate and severe) in brain scans. The stacked ensemble model is able to robustly predict rigid-body motion severity across different acquisition parameters, including T1-weighted and T2-weighted slices acquired in different anatomical planes. The ensemble model with XGBoost metalearner achieves 91.6% accuracy, 94.8% area under the curve, 90% Cohen's Kappa, and is observed to be more accurate and robust than the individual base learners.

An Automated Method for Quality Control in MRI Systems: Methods and Considerations

Objective: The purpose of this study was to develop an automated method for performing quality control (QC) tests in magnetic resonance imaging (MRI) systems, investigate the effect of different definitions of QC parameters and its sensitivity with respect to variations in regions of interest (ROI) positioning, and validate the reliability of the automated method by comparison with results from manual evaluations. Materials and Methods: Magnetic Resonance imaging MRI used for acceptance and routine QC tests from five MRI systems were selected. All QC tests were performed using the American College of Radiology (ACR) MRI accreditation phantom. The only selection criterion was that in the same QC test, images from two identical sequential sequences should be available. The study was focused on four QC parameters: percent signal ghosting (PSG), percent image uniformity (PIU), signal-to-noise ratio (SNR), and SNR uniformity (SNRU), whose values are calculated using the mean signal and the standard deviation of ROIs defined within the phantom image or in the background. The variability of manual ROIs placement was emulated by the software using random variables that follow appropriate normal distributions. Results: Twenty-one paired sequences were employed. The automated test results for PIU were in good agreement with manual results. However, the PSG values were found to vary depending on the selection of ROIs with respect to the phantom. The values of SNR and SNRU also vary significantly, depending on the combination of the two out of the four standard rectangular ROIs. Furthermore, the methodology used for SNR and SNRU calculation also had significant effect on the results. Conclusions: The automated method standardizes the position of ROIs with respect to the ACR phantom image and allows for reproducible QC results.

Sensitivity analysis of different quality assurance methods for magnetic resonance imaging in radiotherapy

BACKGROUND AND PURPOSE

There are currently no standard quality assurance (QA) methods for magnetic resonance imaging (MRI) in radiotherapy (RT). This work was aimed at evaluating the ability of two QA protocols to detect common events that affect quality of MR images under RT settings.

MATERIALS AND METHODS

The American College of Radiology (ACR) MRI QA phantom was repeatedly scanned using a flexible coil and action limits for key image quality parameters were derived. Using an exploratory survey, issues that reduce MR image quality were identified. The most commonly occurring events were introduced as provocations to produce MR images with degraded quality. From these images, detection sensitivities of the ACR MRI QA protocol and a commercial geometric accuracy phantom were determined.

RESULTS

Machine-specific action limits for key image quality parameters set at mean ± 3 σ were comparable with the ACR acceptable values. For the geometric accuracy phantom, provocations from uncorrected gradient nonlinearity effects and a piece of metal in the bore of the scanner resulted in worst distortions of 22.2 mm and 3.4 mm, respectively. The ACR phantom was sensitive to uncorrected signal variations, electric interference and a piece of metal in the bore of the scanner but could not adequately detect individual coil element failures.

CONCLUSIONS

The ACR MRI QA phantom combined with the large field-of-view commercial geometric accuracy phantom were generally sensitive in identifying some common MR image quality issues. The two protocols when combined may provide a tool to monitor the performance of MRI systems in the radiotherapy environment.

Reproducibility of Structural and Diffusion Tensor Imaging in the TACERN Multi-Center Study

BACKGROUND

Multi-site MRI studies are often necessary for recruiting sufficiently sized samples when studying rare conditions. However, they require pooling data from multiple scanners into a single data set, and therefore it is critical to evaluate the variability of quantitative MRI measures within and across scanners used in multi-site studies. The aim of this study was to evaluate the reproducibility of structural and diffusion weighted (DW) MRI measurements acquired on seven scanners at five medical centers as part of the Tuberous Sclerosis Complex Autism Center of Excellence Research Network (TACERN) multisite study.

METHODS

The American College of Radiology (ACR) phantom was imaged monthly to measure reproducibility of signal intensity and uniformity within and across seven 3T scanners from General Electric, Philips, and Siemens vendors. One healthy adult male volunteer was imaged repeatedly on all seven scanners under the TACERN structural and DW protocol (5 b = 0 s/mm2 and 30 b = 1000 s/mm2) over a period of 5 years (age 22-27 years). Reproducibility of inter- and intra-scanner brain segmentation volumes and diffusion tensor imaging metrics fractional anisotropy (FA) and mean diffusivity (MD) within white matter regions was quantified with coefficient of variation.

RESULTS

The American College of Radiology Phantom signal intensity and uniformity were similar across scanners and changed little over time, with a mean intra-scanner coefficient of variation of 3.6 and 1.8%, respectively. The mean inter- and intra-scanner coefficients of variation of brain structure volumes derived from T1-weighted (T1w) images of the human phantom were 3.3 and 1.1%, respectively. The mean inter- and intra-scanner coefficients of variation of FA in white matter regions were 4.5 and 2.5%, while the mean inter- and intra-scanner coefficients of variation of MD in white matter regions were 5.4 and 1.5%.

CONCLUSION

Our results suggest that volumetric and diffusion tensor imaging (DTI) measurements are highly reproducible between and within scanners and provide typical variation amplitudes that can be used as references to interpret future findings in the TACERN network.

Quality assurance of human functional magnetic resonance imaging: a literature review

Functional magnetic resonance imaging (fMRI) has been a popular approach in brain research over the past 20 years. It offers a noninvasive method to probe the brain and uses blood oxygenation level dependent (BOLD) signal changes to access brain function. However, the BOLD signal only represents a small fraction of the total MR signal. System instability and various noise have a strong impact on the BOLD signal. Additionally, fMRI applies fast imaging technique to record brain cognitive process over time, requiring high temporal stability of MR scanners. Furthermore, data acquisition, image quality, processing, and statistical analysis methods also have a great effect on the results of fMRI studies. Quality assurance (QA) programs for fMRI can test the stability of MR scanners, evaluate the quality of fMRI and help to find errors during fMRI scanning, thereby greatly enhancing the success rate of fMRI. In this review, we focus on previous studies which developed QA programs and methods in SCI/SCIE citation peer-reviewed publications over the last 20 years, including topics on existing fMRI QA programs, QA phantoms, image QA metrics, quality evaluation of existing preprocessing pipelines and fMRI statistical analysis methods. The summarized studies were classified into four categories: QA of fMRI systems, QA of fMRI data, quality evaluation of data processing pipelines and statistical methods and QA of task-related fMRI. Summary tables and figures of QA programs and metrics have been developed based on the comprehensive review of the literature.

Standardization of Small Animal Imaging-Current Status and Future Prospects

The benefit of small animal imaging is directly linked to the validity and reliability of the collected data. If the data (regardless of the modality used) are not reproducible and/or reliable, then the outcome of the data is rather questionable. Therefore, standardization of the use of small animal imaging equipment, as well as of animal handling in general, is of paramount importance. In a recent paper, guidance for efficient small animal imaging quality control was offered and discussed, among others, the use of phantoms in setting up a quality control program (Osborne et al. 2016). The same phantoms can be used to standardize image quality parameters for multi-center studies or multi-scanners within center studies. In animal experiments, the additional complexity due to animal handling needs to be addressed to ensure standardized imaging procedures. In this review, we will address the current status of standardization in preclinical imaging, as well as potential benefits from increased levels of standardization.

Penile bulb sparing in prostate cancer radiotherapy : Dose analysis of an in-house MRI system to improve contouring

OBJECTIVE

This study aimed to assess the reduction in dose to the penile bulb (PB) achieved by MRI-based contouring following drinking and endorectal balloon (ERB) instructions.

PATIENTS AND METHODS

A total of 17 prostate cancer patients were treated with intensity-modulated radiation therapy (IMRT) and interstitial brachytherapy (IBT). CT and MRI datasets were acquired back-to-back based on a 65 cm³ air-filled ERB and drinking instructions. After rigid co-registration of the imaging data, the CT-based planning target volume (PTV) used for treatment planning was retrospectively compared to an MRI-based adaptive PTV and the dose to the PB was determined in each case. The adapted PTV encompassed a caudally cropped CT-based PTV which was defined on the basis of the MRI-based prostate contour plus an additional 5 mm safety margin.

RESULTS

In the seven-field IMRT treatment plans, the MRI-based adapted PTV achieved mean (D_mean) and maximum (D_max) doses to the PB which were significantly lower (by 7.6 Gy and 10.9 Gy, respectively; p <0.05) than those of the CT-contoured PTV. For 6 patients, the estimated PB D_max (seven-field IMRT and IBT) for the adapted PTV was <70 Gy, whereas only 1 patient fulfilled this criterium with the CT-based PTV.

CONCLUSION

MRI-based contouring and seven-field IMRT-based treatment planning achieved dose sparing to the PB. Whereas the comparison of MRI and CT contouring only relates to external beam radiotherapy (EBRT) sparing, considering EBRT and IBT shows the improvement in PB sparing for the total treatment.

[Big data and artificial intelligence for diagnostic decision support in atypical dementia]

The differential diagnosis of atypical dementia remains difficult. The use of positron emission tomography (PET) still represents the gold standard for imaging diagnostics. According to the current evidence, however, magnetic resonance imaging (MRI) is almost equal to fluorodeoxyglucose (FDG)-PET, but only when using new big data and machine learning methods. In cases of atypical dementia, especially in younger patients and for follow-up, MRI is preferable to computed tomography (CT). In the clinical routine, promising MRI procedures are e. g. the automated volumetry of anatomical 3D images, as well as a non-contrast-enhanced MRI perfusion method, called arterial spin labeling (ASL). Because of the rapidly growing amount of biomarker data, there is a need for computer-aided big data analyses and artificial intelligence. Based on fast analyses of the diverse and rapidly increasing amount of clinical, imaging, epidemiological, molecular genetic and economic data, new knowledge on the pathogenesis, prevention and treatment can be generated. Technical availability, homogenization of the underlying data and the availability of large reference data are the basis for the widespread establishment of promising analytical methods.

Design and application of an MR reference phantom for multicentre lung imaging trials

Introduction

As there is an increasing number of multicentre lung imaging studies with MRI in patients, dedicated reference phantoms are required to allow for the assessment and comparison of image quality in multi-vendor and multi-centre environments. However, appropriate phantoms for this purpose are so far not available commercially. It was therefore the purpose of this project to design and apply a cost-effective and simple to use reference phantom which addresses the specific requirements for imaging the lungs with MRI.

Methods

The phantom was designed to simulate 4 compartments (lung, blood, muscle and fat) which reflect the specific conditions in proton-MRI of the chest. Multiple phantom instances were produced and measured at 15 sites using a contemporary proton-MRI protocol designed for an in vivo COPD study at intervals over the course of the study. Measures of signal- and contrast-to-noise ratio, as well as structure and edge depiction were extracted from conventionally acquired images using software written for this purpose.

Results

For the signal to noise ratio, low intra-scanner variability was found with 4.5% in the lung compartment, 4.0% for blood, 3.3% for muscle and 3.7% for fat. The inter-scanner variability was substantially higher, with 41%, 32%, 27% and 32% for the same order of compartments. In addition, measures of structure and edge depiction were found to both vary significantly among several scanner types and among scanners of the same model which were equipped with different gradient systems.

Conclusion

The described reference phantom reproducibly quantified image quality aspects and detected substantial inter-scanner variability in a typical pulmonary multicentre proton MRI study, while variability was greater in lung tissue compared to other tissue types. Accordingly, appropriate reference phantoms can help to detect bias in multicentre in vivo study results and could also be used to harmonize equipment or data.

Information-Based Medicine in Glioma Patients: A Clinical Perspective

Glioma constitutes the most common type of primary brain tumor with a dismal survival, often measured in terms of months or years. The thin line between treatment effectiveness and patient harm underpins the importance of tailoring clinical management to the individual patient. Randomized trials have laid the foundation for many neuro-oncological guidelines. Despite this, their findings focus on group-level estimates. Given our current tools, we are limited in our ability to guide patients on what therapy is best for them as individuals, or even how long they should expect to survive. Machine learning, however, promises to provide the analytical support for personalizing treatment decisions, and deep learning allows clinicians to unlock insight from the vast amount of unstructured data that is collected on glioma patients. Although these novel techniques have achieved astonishing results across a variety of clinical applications, significant hurdles remain associated with the implementation of them in clinical practice. Future challenges include the assembly of well-curated cross-institutional datasets, improvement of the interpretability of machine learning models, and balancing novel evidence-based decision-making with the associated liability of automated inference. Although artificial intelligence already exceeds clinical expertise in a variety of applications, clinicians remain responsible for interpreting the implications of, and acting upon, each prediction.

Considerations for the Use of Neuroimaging Technologies for Predicting Recovery of Speech and Language in Aphasia

PURPOSE

The number of research articles aimed at identifying neuroimaging biomarkers for predicting recovery from aphasia continues to grow. Although the clinical use of these biomarkers to determine prognosis has been proposed, there has been little discussion of how this would be accomplished. This is an important issue because the best translational science occurs when translation is considered early in the research process. The purpose of this clinical focus article is to present a framework to guide the discussion of how neuroimaging biomarkers for recovery from aphasia could be implemented clinically.

METHOD

The genomics literature reveals that implementing genetic testing in the real-world poses both opportunities and challenges. There is much similarity between these opportunities and challenges and those related to implementing neuroimaging testing to predict recovery in aphasia. Therefore, the Center for Disease Control's model list of questions aimed at guiding the review of genetic testing has been adapted to guide the discussion of using neuroimaging biomarkers as predictors of recovery in aphasia.

CONCLUSION

The adapted model list presented here is a first and useful step toward initiating a discussion of how neuroimaging biomarkers of recovery could be employed clinically to provide improved quality of care for individuals with aphasia.

Design and first baseline data of the DZNE multicenter observational study on predementia Alzheimer's disease (DELCODE)

BACKGROUND

Deep phenotyping and longitudinal assessment of predementia at-risk states of Alzheimer's disease (AD) are required to define populations and outcomes for dementia prevention trials. Subjective cognitive decline (SCD) is a pre-mild cognitive impairment (pre-MCI) at-risk state of dementia, which emerges as a highly promising target for AD prevention.

METHODS

The German Center for Neurodegenerative Diseases (DZNE) is conducting the multicenter DZNE-Longitudinal Cognitive Impairment and Dementia Study (DELCODE), which focuses on the characterization of SCD in patients recruited from memory clinics. In addition, individuals with amnestic MCI, mild Alzheimer's dementia patients, first-degree relatives of patients with Alzheimer's dementia, and cognitively unimpaired control subjects are studied. The total number of subjects to be enrolled is 1000. Participants receive extensive clinical and neuropsychological assessments, magnetic resonance imaging, positron emission tomography, and biomaterial collection is perfomed. In this publication, we report cognitive and clinical data as well as apolipoprotein E (APOE) genotype and cerebrospinal fluid (CSF) biomarker results of the first 394 baseline data sets.

RESULTS

In comparison with the control group, patients with SCD showed slightly poorer performance on cognitive and functional measures (Alzheimer's Disease Assessment Scale-cognitive part, Clinical Dementia Rating, Functional Activities Questionnaire), with all mean scores in a range which would be considered unimpaired. APOE4 genotype was enriched in the SCD group in comparison to what would be expected in the population and the frequency was significantly higher in comparison to the control group. CSF Aβ42 was lower in the SCD group in comparison to the control group at a statistical trend with age as a covariate. There were no group differences in Tau or pTau concentrations between the SCD and the control groups. The differences in all measures between the MCI group and the AD group were as expected.

CONCLUSIONS

The initial baseline data for DELCODE support the approach of using SCD in patients recruited through memory clinics as an enrichment strategy for late-stage preclinical AD. This is indicated by slightly lower performance in a range of measures in SCD in comparison to the control subjects as well as by enriched APOE4 frequency and lower CSF Aβ42 concentration.

TRIAL REGISTRATION

German Clinical Trials Register DRKS00007966 . Registered 4 May 2015.

Automated image quality evaluation of T2 -weighted liver MRI utilizing deep learning architecture

PURPOSE

To develop and test a deep learning approach named Convolutional Neural Network (CNN) for automated screening of T₂ -weighted (T₂ WI) liver acquisitions for nondiagnostic images, and compare this automated approach to evaluation by two radiologists.

MATERIALS AND METHODS

We evaluated 522 liver magnetic resonance imaging (MRI) exams performed at 1.5T and 3T at our institution between November 2014 and May 2016 for CNN training and validation. The CNN consisted of an input layer, convolutional layer, fully connected layer, and output layer. 351 T₂ WI were anonymized for training. Each case was annotated with a label of being diagnostic or nondiagnostic for detecting lesions and assessing liver morphology. Another independently collected 171 cases were sequestered for a blind test. These 171 T₂ WI were assessed independently by two radiologists and annotated as being diagnostic or nondiagnostic. These 171 T₂ WI were presented to the CNN algorithm and image quality (IQ) output of the algorithm was compared to that of two radiologists.

RESULTS

There was concordance in IQ label between Reader 1 and CNN in 79% of cases and between Reader 2 and CNN in 73%. The sensitivity and the specificity of the CNN algorithm in identifying nondiagnostic IQ was 67% and 81% with respect to Reader 1 and 47% and 80% with respect to Reader 2. The negative predictive value of the algorithm for identifying nondiagnostic IQ was 94% and 86% (relative to Readers 1 and 2).

CONCLUSION

We demonstrate a CNN algorithm that yields a high negative predictive value when screening for nondiagnostic T₂ WI of the liver.

LEVEL OF EVIDENCE

2 Technical Efficacy: Stage 2 J. Magn. Reson. Imaging 2018;47:723-728.

Automated low-contrast pattern recognition algorithm for magnetic resonance image quality assessment

PURPOSE

Low contrast (LC) detectability is a common test criterion for diagnostic radiologic quality control (QC) programs. Automation of this test is desirable in order to reduce human variability and to speed up analysis. However, automation is challenging due to the complexity of the human visual perception system and the ability to create algorithms that mimic this response. This paper describes the development and testing of an automated LC detection algorithm for use in the analysis of magnetic resonance (MR) images of the American College of Radiology (ACR) QC phantom.

METHODS

The detection algorithm includes fuzzy logic decision processes and various edge detection methods to quantify LC detectability. Algorithm performance was first evaluated using a single LC phantom MR image with the addition of incremental zero mean Gaussian noise resulting in a total of 200 images. A c-statistic was calculated to determine the role of CNR to indicate when the algorithm would detect ten spokes. To evaluate inter-rater agreement between experienced observers and the algorithm, a blinded observer study was performed on 196 LC phantom images acquired from nine clinical MR scanners. The nine scanners included two MR manufacturers and two field strengths (1.5 T, 3.0 T). Inter-rater and algorithm-rater agreement was quantified using Krippendorff's alpha.

RESULTS

For the Gaussian noise added data, CNR ranged from 0.519 to 11.7 with CNR being considered an excellent discriminator of algorithm performance (c-statistic = 0.9777). Reviewer scoring of the clinical phantom data resulted in an inter-rater agreement of 0.673 with the agreement between observers and algorithm equal to 0.652, both of which indicate significant agreement.

CONCLUSIONS

This study demonstrates that the detection of LC test patterns for MR imaging QC programs can be successfully developed and that their response can model the human visual detection system of expert MR QC readers.

Image quality assessment of a 1.5T dedicated magnetic resonance-simulator for radiotherapy with a flexible radio frequency coil setting using the standard American College of Radiology magnetic resonance imaging phantom test

BACKGROUND

A flexible RF coil setting has to be used on an MR-simulator (MR-sim) in the head and neck simulation scan for radiotherapy (RT) purpose, while the image quality might be compromised due to the sub-optimized flexible coil compared to the normal diagnostic radiological (DR) head coil. In this study, we assessed the image quality of an MR-sim by conducting the standard American College of Radiology (ACR) MRI phantom test on a 1.5T MR-sim under RT-setting and comparing it to DR-setting.

METHODS

A large ACR MRI phantom was carefully positioned, aligned and scanned 9 times for each under RT- and DR-setting on a 1.5T MR-sim, following the ACR scanning instruction. Images were analyzed following the ACR guidance. Measurement results under two coil settings were quantitatively compared. Inter-observer disagreements under RT-setting between two physicists were compared using Bland-Altman (BA) analysis and intra-class correlation coefficient (ICC).

RESULTS

The MR-sim with RT-setting obtained sufficiently good image quality to pass all ACR recommended criteria. No significant difference was found in phantom length accuracy, high-contrast spatial resolution, slice thickness accuracy, slice position accuracy, and percent-signal ghosting. RT-setting significantly under-performed in low-contrast object detectability, while better performed in image intensity uniformity. BA analysis showed that 95% limit of agreement and biases of phantom test measurement under RT-setting between two observers were very small. Excellent inter-observer agreement (ICC >0.75) was achieved in all measurements except for slice thickness accuracy (ICC =0.42, moderate agreement) under RT-setting.

CONCLUSIONS

Very good and highly reproducible image quality could be achieved on a 1.5T MR-sim with a flexible coil setting as revealed by the standard ACR MRI phantom test. The flexible RT-setting compromised in image signal-to-noise ratio (SNR) compared to the normal DR-setting, and resulted in reduced low-contrast object detectability.

Technical Note: Quantitative dynamic contrast-enhanced MRI of a 3-dimensional artificial capillary network

PURPOSE

Variability across devices, patients, and time still hinders widespread recognition of dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) as quantitative biomarker. The purpose of this work was to introduce and characterize a dedicated microchannel phantom as a model for quantitative DCE-MRI measurements.

METHODS

A perfusable, MR-compatible microchannel network was constructed on the basis of sacrificial melt-spun sugar fibers embedded in a block of epoxy resin. Structural analysis was performed on the basis of light microscopy images before DCE-MRI experiments. During dynamic acquisition the capillary network was perfused with a standard contrast agent injection system. Flow-dependency, as well as inter- and intrascanner reproducibility of the computed DCE parameters were evaluated using a 3.0 T whole-body MRI.

RESULTS

Semi-quantitative and quantitative flow-related parameters exhibited the expected proportionality to the set flow rate (mean Pearson correlation coefficient: 0.991, P < 2.5e-5). The volume fraction was approximately independent from changes of the applied flow rate through the phantom. Repeatability and reproducibility experiments yielded maximum intrascanner coefficients of variation (CV) of 4.6% for quantitative parameters. All evaluated parameters were well in the range of known in vivo results for the applied flow rates.

CONCLUSION

The constructed phantom enables reproducible, flow-dependent, contrast-enhanced MR measurements with the potential to facilitate standardization and comparability of DCE-MRI examinations.

Minimizing noise in pediatric task-based functional MRI; Adolescents with developmental disabilities and typical development

Functional Magnetic Resonance Imaging (fMRI) represents a powerful tool with which to examine brain functioning and development in typically developing pediatric groups as well as children and adolescents with clinical disorders. However, fMRI data can be highly susceptible to misinterpretation due to the effects of excessive levels of noise, often related to head motion. Imaging children, especially with developmental disorders, requires extra considerations related to hyperactivity, anxiety and the ability to perform and maintain attention to the fMRI paradigm. We discuss a number of methods that can be employed to minimize noise, in particular movement-related noise. To this end we focus on strategies prior to, during and following the data acquisition phase employed primarily within our own laboratory. We discuss the impact of factors such as experimental design, screening of potential participants and pre-scan training on head motion in our adolescents with developmental disorders and typical development. We make some suggestions that may minimize noise during data acquisition itself and finally we briefly discuss some current processing techniques that may help to identify and remove noise in the data. Many advances have been made in the field of pediatric imaging, particularly with regard to research involving children with developmental disorders. Mindfulness of issues such as those discussed here will ensure continued progress and greater consistency across studies.

[Quantitative perfusion imaging in magnetic resonance imaging]

CLINICAL/METHODICAL ISSUE

Magnetic resonance imaging (MRI) is recognized for its superior tissue contrast while being non-invasive and free of ionizing radiation. Due to the development of new scanner hardware and fast imaging techniques during the last decades, access to tissue and organ functions became possible. One of these functional imaging techniques is perfusion imaging with which tissue perfusion and capillary permeability can be determined from dynamic imaging data.

STANDARD RADIOLOGICAL METHODS

Perfusion imaging by MRI can be performed by two approaches, arterial spin labeling (ASL) and dynamic contrast-enhanced (DCE) MRI. While the first method uses magnetically labelled water protons in arterial blood as an endogenous tracer, the latter involves the injection of a contrast agent, usually gadolinium (Gd), as a tracer for calculating hemodynamic parameters.

PERFORMANCE

Studies have demonstrated the potential of perfusion MRI for diagnostics and also for therapy monitoring.

ACHIEVEMENTS

The utilization and application of perfusion MRI are still restricted to specialized centers, such as university hospitals. A broad application of the technique has not yet been implemented.

PRACTICAL RECOMMENDATIONS

The MRI perfusion technique is a valuable tool that might come broadly available after implementation of standards on European and international levels. Such efforts are being promoted by the respective professional bodies.

Design of a multimodal (1H/23Na MR/CT) anthropomorphic thorax phantom

OBJECTIVES

This work proposes a modular, anthropomorphic MR and CT thorax phantom that enables the comparison of experimental studies for quantitative evaluation of deformable, multimodal image registration algorithms and realistic multi-nuclear MR imaging techniques.

METHODS

A human thorax phantom was developed with insertable modules representing lung, liver, ribs and additional tracking spheres. The quality of human tissue mimicking characteristics was evaluated for ¹H and ²³Na MR as well as CT imaging. The position of landmarks in the lung lobes was tracked during CT image acquisition at several positions during breathing cycles. ¹H MR measurements of the liver were repeated after seven months to determine long term stability.

RESULTS

The modules possess HU, T₁ and T₂ values comparable to human tissues (lung module: -756±148HU, artificial ribs: 218±56HU (low CaCO₃ concentration) and 339±121 (high CaCO₃ concentration), liver module: T₁=790±28ms, T₂=65±1ms). Motion analysis showed that the landmarks in the lung lobes follow a 3D trajectory similar to human breathing motion. The tracking spheres are well detectable in both CT and MRI. The parameters of the tracking spheres can be adjusted in the following ranges to result in a distinct signal: HU values from 150 to 900HU, T₁ relaxation time from 550ms to 2000ms, T₂ relaxation time from 40ms to 200ms.

CONCLUSION

The presented anthropomorphic multimodal thorax phantom fulfills the demands of a simple, inexpensive system with interchangeable components. In future, the modular design allows for complementing the present set up with additional modules focusing on specific research targets such as perfusion studies, ²³Na MR quantification experiments and an increasing level of complexity for motion studies.

Reproducibility and variability of quantitative magnetic resonance imaging markers in cerebral small vessel disease

Brain imaging is essential for the diagnosis and characterization of cerebral small vessel disease. Several magnetic resonance imaging markers have therefore emerged, providing new information on the diagnosis, progression, and mechanisms of small vessel disease. Yet, the reproducibility of these small vessel disease markers has received little attention despite being widely used in cross-sectional and longitudinal studies. This review focuses on the main small vessel disease-related markers on magnetic resonance imaging including: white matter hyperintensities, lacunes, dilated perivascular spaces, microbleeds, and brain volume. The aim is to summarize, for each marker, what is currently known about: (1) its reproducibility in studies with a scan-rescan procedure either in single or multicenter settings; (2) the acquisition-related sources of variability; and, (3) the techniques used to minimize this variability. Based on the results, we discuss technical and other challenges that need to be overcome in order for these markers to be reliably used as outcome measures in future clinical trials. We also highlight the key points that need to be considered when designing multicenter magnetic resonance imaging studies of small vessel disease.

A connectivity-based test-retest dataset of multi-modal magnetic resonance imaging in young healthy adults

Recently, magnetic resonance imaging (MRI) has been widely used to investigate the structures and functions of the human brain in health and disease in vivo. However, there are growing concerns about the test-retest reliability of structural and functional measurements derived from MRI data. Here, we present a test-retest dataset of multi-modal MRI including structural MRI (S-MRI), diffusion MRI (D-MRI) and resting-state functional MRI (R-fMRI). Fifty-seven healthy young adults (age range: 19-30 years) were recruited and completed two multi-modal MRI scan sessions at an interval of approximately 6 weeks. Each scan session included R-fMRI, S-MRI and D-MRI data. Additionally, there were two separated R-fMRI scans at the beginning and at the end of the first session (approximately 20 min apart). This multi-modal MRI dataset not only provides excellent opportunities to investigate the short- and long-term test-retest reliability of the brain's structural and functional measurements at the regional, connectional and network levels, but also allows probing the test-retest reliability of structural-functional couplings in the human brain.

On replacing the manual measurement of ACR phantom images performed by MRI technologists with an automated measurement approach

PURPOSE

To assess whether measurements on American College of Radiology (ACR) phantom images performed by magnetic resonance imaging (MRI) technologists as part of a weekly quality control (QC) program could be performed exclusively using an automated system without compromising the integrity of the QC program.

MATERIALS AND METHODS

ACR phantom images are acquired on 15 MRI scanners at a number of ACR-accredited sites to fulfill requirements of a weekly QC program. MRI technologists routinely perform several measurements on these images. Software routines are also used to perform the measurements. A set of geometry measurements made by technologists over a five week period and those made using software routines were compared to reference-standard measurements made by two MRI physicists.

RESULTS

The geometry measurements performed by software routines had a very high positive correlation (0.92) with the reference-standard measurements. Technologist measurements also had a high positive correlation (0.63), although the correlation was less than for the automated measurements. Bland-Altman analysis revealed overall good agreement between the automated and reference-standard measurements, with the 95% limits of agreement being within ±0.62 mm. Agreement between the technologist and the reference-standard measurements was demonstratively poorer, with 95% limits of agreement being ±1.46 mm. Some of the technologist measurements differed from the reference standard by as much as 2 mm.

CONCLUSION

The technologists' geometry measurements may be able to be replaced by automated measurement without compromising the weekly QC program required by the ACR.

Comparison of Automated Brain Volume Measures obtained with NeuroQuant and FreeSurfer

PURPOSE

To examine intermethod reliabilities and differences between FreeSurfer and the FDA-cleared congener, NeuroQuant, both fully automated methods for structural brain MRI measurements.

MATERIALS AND METHODS

MRI scans from 20 normal control subjects, 20 Alzheimer's disease patients, and 20 mild traumatically brain-injured patients were analyzed with NeuroQuant and with FreeSurfer. Intermethod reliability was evaluated.

RESULTS

Pairwise correlation coefficients, intraclass correlation coefficients, and effect size differences were computed. NeuroQuant versus FreeSurfer measures showed excellent to good intermethod reliability for the 21 regions evaluated (r: .63 to .99/ICC: .62 to .99/ES: -.33 to 2.08) except for the pallidum (r/ICC/ES = .31/.29/-2.2) and cerebellar white matter (r/ICC/ES = .31/.31/.08). Volumes reported by NeuroQuant were generally larger than those reported by FreeSurfer with the whole brain parenchyma volume reported by NeuroQuant 6.50% larger than the volume reported by FreeSurfer. There was no systematic difference in results between the 3 subgroups.

CONCLUSION

NeuroQuant and FreeSurfer showed good to excellent intermethod reliability in volumetric measurements for all brain regions examined with the only exceptions being the pallidum and cerebellar white matter. This finding was robust for normal individuals, patients with Alzheimer's disease, and patients with mild traumatic brain injury.

An open science resource for establishing reliability and reproducibility in functional connectomics

Efforts to identify meaningful functional imaging-based biomarkers are limited by the ability to reliably characterize inter-individual differences in human brain function. Although a growing number of connectomics-based measures are reported to have moderate to high test-retest reliability, the variability in data acquisition, experimental designs, and analytic methods precludes the ability to generalize results. The Consortium for Reliability and Reproducibility (CoRR) is working to address this challenge and establish test-retest reliability as a minimum standard for methods development in functional connectomics. Specifically, CoRR has aggregated 1,629 typical individuals’ resting state fMRI (rfMRI) data (5,093 rfMRI scans) from 18 international sites, and is openly sharing them via the International Data-sharing Neuroimaging Initiative (INDI). To allow researchers to generate various estimates of reliability and reproducibility, a variety of data acquisition procedures and experimental designs are included. Similarly, to enable users to assess the impact of commonly encountered artifacts (for example, motion) on characterizations of inter-individual variation, datasets of varying quality are included.