MR Artifacts, Safety, and Quality Control

The effect of head motion on brain age prediction using deep convolutional neural networks

Deep learning can be used effectively to predict participants' age from brain magnetic resonance imaging (MRI) data, and a growing body of evidence suggests that the difference between predicted and chronological age-referred to as brain-predicted age difference (brain-PAD)-is related to various neurological and neuropsychiatric disease states. A crucial aspect of the applicability of brain-PAD as a biomarker of individual brain health is whether and how brain-predicted age is affected by MR image artifacts commonly encountered in clinical settings. To investigate this issue, we trained and validated two different 3D convolutional neural network architectures (CNNs) from scratch and tested the models on a separate dataset consisting of motion-free and motion-corrupted T1-weighted MRI scans from the same participants, the quality of which were rated by neuroradiologists from a clinical diagnostic point of view. Our results revealed a systematic increase in brain-PAD with worsening image quality for both models. This effect was also observed for images that were deemed usable from a clinical perspective, with brains appearing older in medium than in good quality images. These findings were also supported by significant associations found between the brain-PAD and standard image quality metrics indicating larger brain-PAD for lower-quality images. Our results demonstrate a spurious effect of advanced brain aging as a result of head motion and underline the importance of controlling for image quality when using brain-predicted age based on structural neuroimaging data as a proxy measure for brain health.

Strategies for improving image quality in prostate MRI

Prostate magnetic resonance imaging (MRI) stands as the cornerstone in diagnosing prostate cancer (PCa), offering superior detection capabilities while minimizing unnecessary biopsies. Despite its critical role, global disparities in MRI diagnostic performance persist, stemming from variations in image quality and radiologist expertise. This manuscript reviews the challenges and strategies for enhancing image quality in prostate MRI, spanning patient preparation, MRI unit optimization, and radiology team engagement. Quality assurance (QA) and quality control (QC) processes are pivotal, emphasizing standardized protocols, meticulous patient evaluation, MRI unit workflow, and radiology team performance. Additionally, artificial intelligence (AI) advancements offer promising avenues for improving image quality and reducing acquisition times. The Prostate-Imaging Quality (PI-QUAL) scoring system emerges as a valuable tool for assessing MRI image quality. A comprehensive approach addressing technical, procedural, and interpretative aspects is essential to ensure consistent and reliable prostate MRI outcomes.

Impaired Glymphatic and Meningeal Lymphatic Functions in Patients with Chronic Migraine

OBJECTIVE

This study was undertaken to investigate migraine glymphatic and meningeal lymphatic vessel (mLV) functions.

METHODS

Migraine patients and healthy controls (HCs) were prospectively recruited between 2020 and 2023. Diffusion tensor image analysis along the perivascular space (DTI-ALPS) index for glymphatics and dynamic contrast-enhanced magnetic resonance imaging parameters (time to peak [TTP]/enhancement integral [EI]/mean time to enhance [MTE]) for para-superior sagittal (paraSSS)-mLV or paratransverse sinus (paraTS)-mLV in episodic migraine (EM), chronic migraine (CM), and CM with and without medication-overuse headache (MOH) were analyzed. DTI-ALPS correlations with clinical parameters (migraine severity [numeric rating scale]/disability [Migraine Disability Assessment (MIDAS)]/bodily pain [Widespread Pain Index]/sleep quality [Pittsburgh Sleep Quality Index (PSQI)]) were examined.

RESULTS

In total, 175 subjects (112 migraine + 63 HCs) were investigated. DTI-ALPS values were lower in CM (median [interquartile range] = 0.64 [0.12]) than in EM (0.71 [0.13], p = 0.005) and HCs (0.71 [0.09], p = 0.004). CM with MOH (0.63 [0.07]) had lower DTI-ALPS values than CM without MOH (0.73 [0.12], p < 0.001). Furthermore, CM had longer TTP (paraSSS-mLV: 55.8 [12.9] vs 40.0 [7.6], p < 0.001; paraTS-mLV: 51.2 [8.1] vs 44.0 [3.3], p = 0.002), EI (paraSSS-mLV: 45.5 [42.0] vs 16.1 [9.2], p < 0.001), and MTE (paraSSS-mLV: 253.7 [6.7] vs 248.4 [13.8], p < 0.001; paraTS-mLV: 252.0 [6.2] vs 249.7 [1.2], p < 0.001) than EM patients. The MIDAS (p = 0.002) and PSQI (p = 0.002) were negatively correlated with DTI-ALPS index after Bonferroni corrections (p < q = 0.01).

INTERPRETATION

CM patients, particularly those with MOH, have glymphatic and meningeal lymphatic dysfunctions, which are highly clinically relevant and may implicate pathogenesis for migraine chronification. ANN NEUROL 2024;95:583-595.

Single Breath-Hold MR Elastography for Fast Biomechanical Probing of Pancreatic Stiffness

BACKGROUND

Pancreatic ductal adenocarcinoma (PDAC) stromal disposition is thought to influence chemotherapy efficacy and increase tissue stiffness, which could be quantified noninvasively via MR elastography (MRE). Current methods cause position-based errors in pancreas location over time, hampering accuracy. It would be beneficial to have a single breath-hold acquisition.

PURPOSE

To develop and test a single breath-hold three-dimensional MRE technique utilizing prospective undersampling and a compressed sensing reconstruction (CS-MRE).

STUDY TYPE

Prospective.

POPULATION

A total of 30 healthy volunteers (HV) (31 ± 9 years; 33% male) and five patients with PDAC (69 ± 5 years; 80% male).

FIELD STRENGTH/SEQUENCE

3-T, GRE Ristretto MRE.

ASSESSMENT

First, optimization of multi breath-hold MRE was done in 10 HV using four combinations of vibration frequency, number of measured wave-phase offsets, and TE and looking at MRE quality measures in the pancreas head. Second, viscoelastic parameters delineated in the pancreas head or tumor of CS-MRE were compared against (I) 2D and (II) 3D four breath-hold acquisitions in HV (N = 20) and PDAC patients. Intrasession repeatability was assessed for CS-MRE in a subgroup of healthy volunteers (N = 15).

STATISTICAL TESTS

Tests include repeated measures analysis of variance (ANOVA), Bland-Altman analysis, and coefficients of variation (CoVs). A P-value <.05 was considered statistically significant.

RESULTS

Optimization of the four breath-hold acquisitions resulted in 40 Hz vibration frequency, five wave-phases, and echo time (TE) = 6.9 msec as the preferred method (4BH-MRE). CS-MRE quantitative results did not differ from 4BH-MRE. Shear wave speed (SWS) and phase angle differed significantly between HV and PDAC patients using 4BH-MRE or CS-MRE. The limits of agreement for SWS were [-0.09, 0.10] m/second and the within-subject CoV was 4.8% for CS-MRE.

DATA CONCLUSION

CS-MRE might allow a single breath-hold MRE acquisition with comparable SWS and phase angle as 4BH-MRE, and it may still enable to differentiate between HV and PDAC.

LEVEL OF EVIDENCE

2 Technical Efficacy Stage: 2.

Update on Optimization of Prostate MR Imaging Technique and Image Quality

Prostate MR imaging quality has improved dramatically over recent times, driven by advances in hardware, software, and improved functional imaging techniques. MRI now plays a key role in prostate cancer diagnostic work-up, but outcomes of the MRI-directed pathway are heavily dependent on image quality and optimization. MR sequences can be affected by patient-related degradations relating to motion and susceptibility artifacts which may enable only partial mitigation. In this Review, we explore issues relating to prostate MRI acquisition and interpretation, mitigation strategies at a patient and scanner level, PI-QUAL reporting, and future directions in image quality, including artificial intelligence solutions.

A Longitudinal MRI-Based Artificial Intelligence System to Predict Pathological Complete Response After Neoadjuvant Therapy in Rectal Cancer: A Multicenter Validation Study

BACKGROUND

Accurate prediction of response to neoadjuvant chemoradiotherapy is critical for subsequent treatment decisions for patients with locally advanced rectal cancer.

OBJECTIVE

To develop and validate a deep learning model based on the comparison of paired MRI before and after neoadjuvant chemoradiotherapy to predict pathological complete response.

DESIGN

By capturing the changes from MRI before and after neoadjuvant chemoradiotherapy in 638 patients, we trained a multitask deep learning model for response prediction (DeepRP-RC) that also allowed simultaneous segmentation. Its performance was independently tested in an internal and 3 external validation sets, and its prognostic value was also evaluated.

SETTINGS

Multicenter study.

PATIENTS

We retrospectively enrolled 1201 patients diagnosed with locally advanced rectal cancer who underwent neoadjuvant chemoradiotherapy before total mesorectal excision. Patients had been treated at 1 of 4 hospitals in China between January 2013 and December 2020.

MAIN OUTCOME MEASURES

The main outcome was the accuracy of predicting pathological complete response, measured as the area under receiver operating curve for the training and validation data sets.

RESULTS

DeepRP-RC achieved high performance in predicting pathological complete response after neoadjuvant chemoradiotherapy, with area under the curve values of 0.969 (0.942-0.996), 0.946 (0.915-0.977), 0.943 (0.888-0.998), and 0.919 (0.840-0.997) for the internal and 3 external validation sets, respectively. DeepRP-RC performed similarly well in the subgroups defined by receipt of radiotherapy, tumor location, T/N stages before and after neoadjuvant chemoradiotherapy, and age. Compared with experienced radiologists, the model showed substantially higher performance in pathological complete response prediction. The model was also highly accurate in identifying the patients with poor response. Furthermore, the model was significantly associated with disease-free survival independent of clinicopathological variables.

LIMITATIONS

This study was limited by its retrospective design and absence of multiethnic data.

CONCLUSIONS

DeepRP-RC could be an accurate preoperative tool for pathological complete response prediction in rectal cancer after neoadjuvant chemoradiotherapy.

UN SISTEMA DE IA BASADO EN RESONANCIA MAGNTICA LONGITUDINAL PARA PREDECIR LA RESPUESTA PATOLGICA COMPLETA DESPUS DE LA TERAPIA NEOADYUVANTE EN EL CNCER DE RECTO UN ESTUDIO DE VALIDACIN MULTICNTRICO

ANTECEDENTES:La predicción precisa de la respuesta a la quimiorradioterapia neoadyuvante es fundamental para las decisiones de tratamiento posteriores para los pacientes con cáncer de recto localmente avanzado.OBJETIVO:Desarrollar y validar un modelo de aprendizaje profundo basado en la comparación de resonancias magnéticas pareadas antes y después de la quimiorradioterapia neoadyuvante para predecir la respuesta patológica completa.DISEÑO:Al capturar los cambios de las imágenes de resonancia magnética antes y después de la quimiorradioterapia neoadyuvante en 638 pacientes, entrenamos un modelo de aprendizaje profundo multitarea para la predicción de respuesta (DeepRP-RC) que también permitió la segmentación simultánea. Su rendimiento se probó de forma independiente en un conjunto de validación interna y tres externas, y también se evaluó su valor pronóstico.ESCENARIO:Estudio multicéntrico.PACIENTES:Volvimos a incluir retrospectivamente a 1201 pacientes diagnosticados con cáncer de recto localmente avanzado y sometidos a quimiorradioterapia neoadyuvante antes de la escisión total del mesorrecto. Eran de cuatro hospitales en China en el período entre enero de 2013 y diciembre de 2020.PRINCIPALES MEDIDAS DE RESULTADO:Los principales resultados fueron la precisión de la predicción de la respuesta patológica completa, medida como el área bajo la curva operativa del receptor para los conjuntos de datos de entrenamiento y validación.RESULTADOS:DeepRP-RC logró un alto rendimiento en la predicción de la respuesta patológica completa después de la quimiorradioterapia neoadyuvante, con valores de área bajo la curva de 0,969 (0,942-0,996), 0,946 (0,915-0,977), 0,943 (0,888-0,998), y 0,919 (0,840-0,997) para los conjuntos de validación interna y las tres externas, respectivamente. DeepRP-RC se desempeñó de manera similar en los subgrupos definidos por la recepción de radioterapia, la ubicación del tumor, los estadios T/N antes y después de la quimiorradioterapia neoadyuvante y la edad. En comparación con los radiólogos experimentados, el modelo mostró un rendimiento sustancialmente mayor en la predicción de la respuesta patológica completa. El modelo también fue muy preciso en la identificación de los pacientes con mala respuesta. Además, el modelo se asoció significativamente con la supervivencia libre de enfermedad independientemente de las variables clinicopatológicas.LIMITACIONES:Este estudio estuvo limitado por el diseño retrospectivo y la ausencia de datos multiétnicos.CONCLUSIONES:DeepRP-RC podría servir como una herramienta preoperatoria precisa para la predicción de la respuesta patológica completa en el cáncer de recto después de la quimiorradioterapia neoadyuvante. (Traducción-Dr. Felipe Bellolio ).

Prevalence and influencing factors for artifact development in breast MRI-derived maximum intensity projections

BACKGROUND

Magnetic resonance imaging (MRI) provides high diagnostic sensitivity for breast cancer. However, MRI artifacts may impede the diagnostic assessment. This is particularly important when evaluating maximum intensity projections (MIPs), such as in abbreviated MRI (AB-MRI) protocols, because high image quality is desired as a result of fewer sequences being available to compensate for problems.

PURPOSE

To describe the prevalence of artifacts on dynamic contrast enhanced (DCE) MRI-derived MIPs and to investigate potentially associated attributes.

MATERIAL AND METHODS

For this institutional review board approved retrospective analysis, MIPs were generated from subtraction series and cropped to represent the left and right breasts as regions of interest. These images were labeled by three independent raters regarding the presence of MRI artifacts. MRI artifact prevalence and associations with patient characteristics and technical attributes were analyzed using descriptive statistics and generalized linear models (GLMMs).

RESULTS

The study included 2524 examinations from 1794 patients (median age 50 years), performed on 1.5 and 3.0 Tesla MRI systems. Overall inter-rater agreement was kappa = 0.54. Prevalence of significant unilateral artifacts was 29.2% (736/2524), whereas bilateral artifacts were present in 37.8% (953/2524) of all examinations. According to the GLMM, artifacts were significantly positive associated with age (odds ratio [OR] = 1.52) and magnetic field strength (OR = 1.55), whereas a negative effect could be shown for body mass index (OR = 0.95).

CONCLUSION

MRI artifacts on DCE subtraction MIPs of the breast, as used in AB-MRI, are a relevant topic. Our results show that, besides the magnetic field strength, further associated attributes are patient age and body mass index, which can provide possible targets for artifact reduction.

Correction of Arterial-Phase Motion Artifacts in Gadoxetic Acid-Enhanced Liver MRI Using an Innovative Unsupervised Network

This study aims to propose and evaluate DR-CycleGAN, a disentangled unsupervised network by introducing a novel content-consistency loss, for removing arterial-phase motion artifacts in gadoxetic acid-enhanced liver MRI examinations. From June 2020 to July 2021, gadoxetic acid-enhanced liver MRI data were retrospectively collected in this center to establish training and testing datasets. Motion artifacts were semi-quantitatively assessed using a five-point Likert scale (1 = no artifact, 2 = mild, 3 = moderate, 4 = severe, and 5 = non-diagnostic) and quantitatively evaluated using the structural similarity index (SSIM) and peak signal-to-noise ratio (PSNR). The datasets comprised a training dataset (308 examinations, including 58 examinations with artifact grade = 1 and 250 examinations with artifact grade ≥ 2), a paired test dataset (320 examinations, including 160 examinations with artifact grade = 1 and paired 160 examinations with simulated motion artifacts of grade ≥ 2), and an unpaired test dataset (474 examinations with artifact grade ranging from 1 to 5). The performance of DR-CycleGAN was evaluated and compared with a state-of-the-art network, Cycle-MedGAN V2.0. As a result, in the paired test dataset, DR-CycleGAN demonstrated significantly higher SSIM and PSNR values and lower motion artifact grades compared to Cycle-MedGAN V2.0 (0.89 ± 0.07 vs. 0.84 ± 0.09, 32.88 ± 2.11 vs. 30.81 ± 2.64, and 2.7 ± 0.7 vs. 3.0 ± 0.9, respectively; p < 0.001 each). In the unpaired test dataset, DR-CycleGAN also exhibited a superior motion artifact correction performance, resulting in a significant decrease in motion artifact grades from 2.9 ± 1.3 to 2.0 ± 0.6 compared to Cycle-MedGAN V2.0 (to 2.4 ± 0.9, p < 0.001). In conclusion, DR-CycleGAN effectively reduces motion artifacts in the arterial phase images of gadoxetic acid-enhanced liver MRI examinations, offering the potential to enhance image quality.

Unsupervised dual-domain disentangled network for removal of rigid motion artifacts in MRI

Motion artifacts in magnetic resonance imaging (MRI) have always been a serious issue because they can affect subsequent diagnosis and treatment. Supervised deep learning methods have been investigated for the removal of motion artifacts; however, they require paired data that are difficult to obtain in clinical settings. Although unsupervised methods are widely proposed to fully use clinical unpaired data, they generally focus on anatomical structures generated by the spatial domain while ignoring phase error (deviations or inaccuracies in phase information that are possibly caused by rigid motion artifacts during image acquisition) provided by the frequency domain. In this study, a 2D unsupervised deep learning method named unsupervised disentangled dual-domain network (UDDN) was proposed to effectively disentangle and remove unwanted rigid motion artifacts from images. In UDDN, a dual-domain encoding module was presented to capture different types of information from the spatial and frequency domains to enrich the information. Moreover, a cross-domain attention fusion module was proposed to effectively fuse information from different domains, reduce information redundancy, and improve the performance of motion artifact removal. UDDN was validated on a publicly available dataset and a clinical dataset. Qualitative and quantitative experimental results showed that our method could effectively remove motion artifacts and reconstruct image details. Moreover, the performance of UDDN surpasses that of several state-of-the-art unsupervised methods and is comparable with that of the supervised method. Therefore, our method has great potential for clinical application in MRI, such as real-time removal of rigid motion artifacts.

Evaluation of Artifact Appearance and Burden in Pediatric Brain Tumor MR Imaging with Compressed Sensing in Comparison to Conventional Parallel Imaging Acceleration

Clinical magnetic resonance imaging (MRI) aims for the highest possible image quality, while balancing the need for acceptable examination time, reasonable signal-to-noise ratio (SNR), and lowest artifact burden. With a recently introduced imaging acceleration technique, compressed sensing, the acquisition speed and image quality of pediatric brain tumor exams can be improved. However, little attention has been paid to its impact on method-related artifacts in pediatric brain MRI. This study assessed the overall artifact burden and artifact appearances in a standardized pediatric brain tumor MRI by comparing conventional parallel imaging acceleration with compressed sensing. This showed that compressed sensing resulted in fewer physiological artifacts in the FLAIR sequence, and a reduction in technical artifacts in the 3D T1 TFE sequences. Only a slight difference was noted in the T2 TSE sequence. A relatively new range of artifacts, which are likely technique-related, was noted in the 3D T1 TFE sequences. In conclusion, by equipping a basic pediatric brain tumor protocol for 3T MRI with compressed sensing, the overall burden of common artifacts can be reduced. However, attention should be paid to novel compressed-sensing-specific artifacts.

ROOD-MRI: Benchmarking the robustness of deep learning segmentation models to out-of-distribution and corrupted data in MRI

Deep artificial neural networks (DNNs) have moved to the forefront of medical image analysis due to their success in classification, segmentation, and detection challenges. A principal challenge in large-scale deployment of DNNs in neuroimage analysis is the potential for shifts in signal-to-noise ratio, contrast, resolution, and presence of artifacts from site to site due to variances in scanners and acquisition protocols. DNNs are famously susceptible to these distribution shifts in computer vision. Currently, there are no benchmarking platforms or frameworks to assess the robustness of new and existing models to specific distribution shifts in MRI, and accessible multi-site benchmarking datasets are still scarce or task-specific. To address these limitations, we propose ROOD-MRI: a novel platform for benchmarking the Robustness of DNNs to Out-Of-Distribution (OOD) data, corruptions, and artifacts in MRI. This flexible platform provides modules for generating benchmarking datasets using transforms that model distribution shifts in MRI, implementations of newly derived benchmarking metrics for image segmentation, and examples for using the methodology with new models and tasks. We apply our methodology to hippocampus, ventricle, and white matter hyperintensity segmentation in several large studies, providing the hippocampus dataset as a publicly available benchmark. By evaluating modern DNNs on these datasets, we demonstrate that they are highly susceptible to distribution shifts and corruptions in MRI. We show that while data augmentation strategies can substantially improve robustness to OOD data for anatomical segmentation tasks, modern DNNs using augmentation still lack robustness in more challenging lesion-based segmentation tasks. We finally benchmark U-Nets and vision transformers, finding robustness susceptibility to particular classes of transforms across architectures. The presented open-source platform enables generating new benchmarking datasets and comparing across models to study model design that results in improved robustness to OOD data and corruptions in MRI.

Automatic brain MRI motion artifact detection based on end-to-end deep learning is similarly effective as traditional machine learning trained on image quality metrics

Head motion artifacts in magnetic resonance imaging (MRI) are an important confounding factor concerning brain research as well as clinical practice. For this reason, several machine learning-based methods have been developed for the automatic quality control of structural MRI scans. Deep learning offers a promising solution to this problem, however, given its data-hungry nature and the scarcity of expert-annotated datasets, its advantage over traditional machine learning methods in identifying motion-corrupted brain scans is yet to be determined. In the present study, we investigated the relative advantage of the two methods in structural MRI quality control. To this end, we collected publicly available T1-weighted images and scanned subjects in our own lab under conventional and active head motion conditions. The quality of the images was rated by a team of radiologists from the point of view of clinical diagnostic use. We present a relatively simple, lightweight 3D convolutional neural network trained in an end-to-end manner that achieved a test set (N = 411) balanced accuracy of 94.41% in classifying brain scans into clinically usable or unusable categories. A support vector machine trained on image quality metrics achieved a balanced accuracy of 88.44% on the same test set. Statistical comparison of the two models yielded no significant difference in terms of confusion matrices, error rates, or receiver operating characteristic curves. Our results suggest that these machine learning methods are similarly effective in identifying severe motion artifacts in brain MRI scans, and underline the efficacy of end-to-end deep learning-based systems in brain MRI quality control, allowing the rapid evaluation of diagnostic utility without the need for elaborate image pre-processing.

Validation of a semi-automatic method to measure total liver volumes in polycystic liver disease on computed tomography - high speed and accuracy

OBJECTIVES

Polycystic liver disease (PLD) is characterized by growth of hepatic cysts, causing hepatomegaly. Disease severity is determined using total liver volume (TLV), which can be measured from computed tomography (CT). The gold standard is manual segmentation which is time-consuming and requires expert knowledge of the anatomy. This study aims to validate the commercially available semi-automatic MMWP (Multimodality Workplace) Volume tool for CT scans of PLD patients.

METHODS

We included adult patients with one (n = 60) or two (n = 46) abdominal CT scans. Semi-automatic contouring was compared with manual segmentation, using comparison of observed volumes (cross-sectional) and growth (longitudinal), correlation coefficients (CC), and Bland-Altman analyses with bias and precision, defined as the mean difference and SD from this difference. Inter- and intra-reader variability were assessed using coefficients of variation (CV) and we assessed the time to perform both procedures.

RESULTS

Median TLV was 5292.2 mL (IQR 3141.4-7862.2 mL) at baseline. Cross-sectional analysis showed high correlation and low bias and precision between both methods (CC 0.998, bias 1.62%, precision 2.75%). Absolute volumes were slightly higher for semi-automatic segmentation (manual 5292.2 (3141.4-7862.2) versus semi-automatic 5432.8 (3071.9-7960.2) mL, difference 2.7%, p < 0.001). Longitudinal analysis demonstrated that semi-automatic segmentation accurately measures liver growth (CC 0.908, bias 0.23%, precision 4.04%). Inter- and intra-reader variability were small (2.19% and 0.66%) and comparable to manual segmentation (1.21% and 0.63%) (p = 0.26 and p = 0.37). Semi-automatic segmentation was faster than manual tracing (19 min versus 50 min, p = 0.009).

CONCLUSIONS

Semi-automatic liver segmentation is a fast and accurate method to determine TLV and liver growth in PLD patients.

KEY POINTS

• Semi-automatic liver segmentation using the commercially available MMWP volume tool accurately determines total liver volume as well as liver growth over time in polycystic liver disease patients. • This method is considerably faster than manual segmentation through the use of Hounsfield unit settings. • We used a real-life CT set for the validation and showed that the semi-automatic tool measures accurately regardless of contrast used for the CT scan or not, presence of polycystic kidneys, liver volume, and previous invasive treatment for polycystic liver disease.

Evaluation of a New, Highly Flexible Radiofrequency Coil for MR Simulation of Patients Undergoing External Beam Radiation Therapy

PURPOSE

To evaluate the performance of a new, highly flexible radiofrequency (RF) coil system for imaging patients undergoing MR simulation.

METHODS

Volumetric phantom and in vivo images were acquired with a commercially available and prototype RF coil set. Phantom evaluation was performed using a silicone-filled humanoid phantom of the head and shoulders. In vivo assessment was performed in five healthy and six patient subjects. Phantom data included T1-weighted volumetric imaging, while in vivo acquisitions included both T1- and T2-weighted volumetric imaging. Signal to noise ratio (SNR) and uniformity metrics were calculated in the phantom data, while SNR values were calculated in vivo. Statistical significance was tested by means of a non-parametric analysis of variance test.

RESULTS

At a threshold of p = 0.05, differences in measured SNR distributions within the entire phantom volume were statistically different in two of the three paired coil set comparisons. Differences in per slice average SNR between the two coil sets were all statistically significant, as well as differences in per slice image uniformity. For patients, SNRs within the entire imaging volume were statistically significantly different in four of the nine comparisons and seven of the nine comparisons performed on the per slice average SNR values. For healthy subjects, SNRs within the entire imaging volume were statistically significantly different in seven of the nine comparisons and eight of the nine comparisons when per slice average SNR was tested.

CONCLUSIONS

Phantom and in vivo results demonstrate that image quality obtained from the novel flexible RF coil set was similar or improved over the conventional coil system. The results also demonstrate that image quality is impacted by the specific coil configurations used for imaging and should be matched appropriately to the anatomic site imaged to ensure optimal and reproducible image quality.

Relaxometry and brain myelin quantification with synthetic MRI in MS subtypes and their associations with spinal cord atrophy

Immune-mediated demyelination and neurodegeneration are pathophysiological hallmarks of Multiple Sclerosis (MS) and main drivers of disease related disability. The principal method for evaluating qualitatively demyelinating events in the clinical context is contrast-weighted magnetic resonance imaging (MRI). Moreover, advanced MRI sequences provide reliable quantification of brain myelin offering new opportunities to study tissue pathology in vivo. Towards neurodegenerative aspects of the disease, spinal cord atrophy - besides brain atrophy - is a powerful and validated predictor of disease progression. The etiology of spinal cord volume loss is still a matter of research, as it remains unclear whether the impact of local lesion pathology or the interaction with supra- and infratentorial axonal degeneration and demyelination of the long descending and ascending fiber tracts are the determining factors. Quantitative synthetic MR using a multiecho acquisition of saturation recovery pulse sequence provides fast automatic brain tissue and myelin volumetry based on R1 and R2 relaxation rates and proton density quantification, making it a promising modality for application in the clinical routine. In this cross sectional study a total of 91 MS patients and 31 control subjects were included to investigate group differences of global and regional measures of brain myelin and relaxation rates, in different MS subtypes, using QRAPMASTER sequence and SyMRI postprocessing software. Furthermore, we examined associations between these quantitative brain parameters and spinal cord atrophy to draw conclusions about possible pathophysiological relationships. Intracranial myelin volume fraction of the global brain exhibited statistically significant differences between control subjects (10.4%) and MS patients (RRMS 9.4%, PMS 8.1%). In a LASSO regression analysis with total brain lesion load, intracranial myelin volume fraction and brain parenchymal fraction, the intracranial myelin volume fraction was the variable with the highest impact on spinal cord atrophy (standardized coefficient 4.52). Regional supratentorial MRI metrics showed altered average myelin volume fraction, R1, R2 and proton density in MS patients compared to controls most pronounced in PMS. Interestingly, quantitative MRI parameters in supratentorial regions showed strong associations with upper cord atrophy, suggesting an important role of brain diffuse demyelination on spinal cord pathology possibly in the context of global disease activity. R1, R2 or proton density of the thalamus, cerebellum and brainstem correlated with upper cervical cord atrophy, probably reflecting the direct functional connection between these brain structures and the spinal cord as well as the effects of retrograde and anterograde axonal degeneration. By using Synthetic MR-derived myelin volume fraction, we were able to effectively detect significant differences of myelination in relapsing and progressive MS subtypes. Total intracranial brain myelin volume fraction seemed to predict spinal cord volume loss better than brain atrophy or total lesion load. Furthermore, demyelination in highly myelinated supratentorial regions, as an indicator of diffuse disease activity, as well as alterations of relaxation parameters in adjacent infratentorial and midbrain areas were strongly associated with upper cervical cord atrophy.

AutoComBat: a generic method for harmonizing MRI-based radiomic features

The use of multicentric data is becoming essential for developing generalizable radiomic signatures. In particular, Magnetic Resonance Imaging (MRI) data used in brain oncology are often heterogeneous in terms of scanners and acquisitions, which significantly impact quantitative radiomic features. Various methods have been proposed to decrease dependency, including methods acting directly on MR images, i.e., based on the application of several preprocessing steps before feature extraction or the ComBat method, which harmonizes radiomic features themselves. The ComBat method used for radiomics may be misleading and presents some limitations, such as the need to know the labels associated with the "batch effect". In addition, a statistically representative sample is required and the applicability of a signature whose batch label is not present in the train set is not possible. This work aimed to compare a priori and a posteriori radiomic harmonization methods and propose a code adaptation to be machine learning compatible. Furthermore, we have developed AutoComBat, which aims to automatically determine the batch labels, using either MRI metadata or quality metrics as inputs of the proposed constrained clustering. A heterogeneous dataset consisting of high and low-grade gliomas coming from eight different centers was considered. The different methods were compared based on their ability to decrease relative standard deviation of radiomic features extracted from white matter and on their performance on a classification task using different machine learning models. ComBat and AutoComBat using image-derived quality metrics as inputs for batch assignment and preprocessing methods presented promising results on white matter harmonization, but with no clear consensus for all MR images. Preprocessing showed the best results on the T1w-gd images for the grading task. For T2w-flair, AutoComBat, using either metadata plus quality metrics or metadata alone as inputs, performs better than the conventional ComBat, highlighting its potential for data harmonization. Our results are MRI weighting, feature class and task dependent and require further investigations on other datasets.

An artificial intelligence multiprocessing scheme for the diagnosis of osteosarcoma MRI images

Osteosarcoma is the most common malignant osteosarcoma, and most developing countries face great challenges in the diagnosis due to the lack of medical resources. Magnetic resonance imaging (MRI) has always been an important tool for the detection of osteosarcoma, but it is a time-consuming and labor-intensive task for doctors to manually identify MRI images. It is highly subjective and prone to misdiagnosis. Existing computer-aided diagnosis methods of osteosarcoma MRI images focus only on accuracy, ignoring the lack of computing resources in developing countries. In addition, the large amount of redundant and noisy data generated during imaging should also be considered. To alleviate the inefficiency of osteosarcoma diagnosis faced by developing countries, this paper proposed an artificial intelligence multiprocessing scheme for pre-screening, noise reduction, and segmentation of osteosarcoma MRI images. For pre-screening, we propose the Slide Block Filter to remove useless images. Next, we introduced a fast non-local means algorithm using integral images to denoise noisy images. We then segmented the filtered and denoised MRI images using a U-shaped network (ETUNet) embedded with a transformer layer, which enhances the functionality and robustness of the traditional U-shaped architecture. Finally, we further optimized the segmented tumor boundaries using conditional random fields. This paper conducted experiments on more than 70,000 MRI images of osteosarcoma from three hospitals in China. The experimental results show that our proposed methods have good results and better performance in pre-screening, noise reduction, and segmentation.

A Brief Survey on No-Reference Image Quality Assessment Methods for Magnetic Resonance Images

No-reference image quality assessment (NR-IQA) methods automatically and objectively predict the perceptual quality of images without access to a reference image. Therefore, due to the lack of pristine images in most medical image acquisition systems, they play a major role in supporting the examination of resulting images and may affect subsequent treatment. Their usage is particularly important in magnetic resonance imaging (MRI) characterized by long acquisition times and a variety of factors that influence the quality of images. In this work, a survey covering recently introduced NR-IQA methods for the assessment of MR images is presented. First, typical distortions are reviewed and then popular NR methods are characterized, taking into account the way in which they describe MR images and create quality models for prediction. The survey also includes protocols used to evaluate the methods and popular benchmark databases. Finally, emerging challenges are outlined along with an indication of the trends towards creating accurate image prediction models.

Simultaneous relaxometry and morphometry of human brain structures with 3D magnetic resonance fingerprinting: a multicenter, multiplatform, multifield-strength study

Relaxation times and morphological information are fundamental magnetic resonance imaging-derived metrics of the human brain that reflect the status of the underlying tissue. Magnetic resonance fingerprinting (MRF) enables simultaneous acquisition of T1 and T2 maps inherently aligned to the anatomy, allowing whole-brain relaxometry and morphometry in a single scan. In this study, we revealed the feasibility of 3D MRF for simultaneous brain structure-wise morphometry and relaxometry. Comprehensive test-retest scan analyses using five 1.5-T and three 3.0-T systems from a single vendor including different scanner types across 3 institutions demonstrated that 3D MRF-derived morphological information and relaxation times are highly repeatable at both 1.5 T and 3.0 T. Regional cortical thickness and subcortical volume values showed high agreement and low bias across different field strengths. The ability to acquire a set of regional T1, T2, thickness, and volume measurements of neuroanatomical structures with high repeatability and reproducibility facilitates the ability of longitudinal multicenter imaging studies to quantitatively monitor changes associated with underlying pathologies, disease progression, and treatments.

Improved Multi-Echo Gradient-Echo-Based Myelin Water Fraction Mapping Using Dimensionality Reduction

Multi-echo gradient-echo (mGRE)-based myelin water fraction (MWF) mapping is a promising myelin water imaging (MWI) modality but is vulnerable to noise and artifact corruption. The linear dimensionality reduction (LDR) method has recently shown improvements with regard to these challenges. However, the magnitude value based low rank operators have been shown to misestimate the MWF for regions with [Formula: see text] anisotropy. This paper presents a nonlinear dimensionality reduction (NLDR) method to estimate the MWF map better by encouraging nonlinear low dimensionality of mGRE signal sources. Specifically, we implemented a fully connected deep autoencoder to extract the low-dimensional features of complex-valued signals and incorporated a sparse regularization to separate the anomaly sources that do not reside in the low-dimensional manifold. Simulations and in vivo experiments were performed to evaluate the accuracy of the MWF map under various situations. The proposed NLDR-based MWF improves the accuracy of the MWF map over the conventional nonlinear least-squares method and the LDR-based MWF and maintains robustness against noise and artifact corruption.

Automated artifact detection in abbreviated dynamic contrast-enhanced (DCE) MRI-derived maximum intensity projections (MIPs) of the breast

OBJECTIVES

To automatically detect MRI artifacts on dynamic contrast-enhanced (DCE) maximum intensity projections (MIPs) of the breast using deep learning.

METHODS

Women who underwent clinically indicated breast MRI between October 2015 and December 2019 were included in this IRB-approved retrospective study. We employed two convolutional neural network architectures (ResNet and DenseNet) to detect the presence of artifacts on DCE MIPs of the left and right breasts. Networks were trained on images acquired up to and including the year 2018 using a 5-fold cross-validation (CV). Ensemble classifiers were built with the resulting CV models and applied to an independent holdout test dataset, which was formed by images acquired in 2019.

RESULTS

Our study sample contained 2265 examinations from 1794 patients (median age at first acquisition: 50 years [IQR: 17 years]), corresponding to 1827 examinations of 1378 individuals in the training dataset and 438 examinations of 416 individuals in the holdout test dataset with a prevalence of image-level artifacts of 53% (1951/3654 images) and 43% (381/876 images), respectively. On the holdout test dataset, the ResNet and DenseNet ensembles demonstrated an area under the ROC curve of 0.92 and 0.94, respectively.

CONCLUSION

Neural networks are able to reliably detect artifacts that may impede the diagnostic assessment of MIPs derived from DCE subtraction series in breast MRI. Future studies need to further explore the potential of such neural networks to complement quality assurance and improve the application of DCE MIPs in a clinical setting, such as abbreviated protocols.

KEY POINTS

• Deep learning classifiers are able to reliably detect MRI artifacts in dynamic contrast-enhanced protocol-derived maximum intensity projections of the breast. • Automated quality assurance of maximum intensity projections of the breast may be of special relevance for abbreviated breast MRI, e.g., in high-throughput settings, such as cancer screening programs.

TorchIO: A Python library for efficient loading, preprocessing, augmentation and patch-based sampling of medical images in deep learning

BACKGROUND AND OBJECTIVE

Processing of medical images such as MRI or CT presents different challenges compared to RGB images typically used in computer vision. These include a lack of labels for large datasets, high computational costs, and the need of metadata to describe the physical properties of voxels. Data augmentation is used to artificially increase the size of the training datasets. Training with image subvolumes or patches decreases the need for computational power. Spatial metadata needs to be carefully taken into account in order to ensure a correct alignment and orientation of volumes.

METHODS

We present TorchIO, an open-source Python library to enable efficient loading, preprocessing, augmentation and patch-based sampling of medical images for deep learning. TorchIO follows the style of PyTorch and integrates standard medical image processing libraries to efficiently process images during training of neural networks. TorchIO transforms can be easily composed, reproduced, traced and extended. Most transforms can be inverted, making the library suitable for test-time augmentation and estimation of aleatoric uncertainty in the context of segmentation. We provide multiple generic preprocessing and augmentation operations as well as simulation of MRI-specific artifacts.

RESULTS

Source code, comprehensive tutorials and extensive documentation for TorchIO can be found at http://torchio.rtfd.io/. The package can be installed from the Python Package Index (PyPI) running pip install torchio. It includes a command-line interface which allows users to apply transforms to image files without using Python. Additionally, we provide a graphical user interface within a TorchIO extension in 3D Slicer to visualize the effects of transforms.

CONCLUSION

TorchIO was developed to help researchers standardize medical image processing pipelines and allow them to focus on the deep learning experiments. It encourages good open-science practices, as it supports experiment reproducibility and is version-controlled so that the software can be cited precisely. Due to its modularity, the library is compatible with other frameworks for deep learning with medical images.

TorchIO: A Python library for efficient loading, preprocessing, augmentation and patch-based sampling of medical images in deep learning

BACKGROUND AND OBJECTIVE

Processing of medical images such as MRI or CT presents different challenges compared to RGB images typically used in computer vision. These include a lack of labels for large datasets, high computational costs, and the need of metadata to describe the physical properties of voxels. Data augmentation is used to artificially increase the size of the training datasets. Training with image subvolumes or patches decreases the need for computational power. Spatial metadata needs to be carefully taken into account in order to ensure a correct alignment and orientation of volumes.

METHODS

We present TorchIO, an open-source Python library to enable efficient loading, preprocessing, augmentation and patch-based sampling of medical images for deep learning. TorchIO follows the style of PyTorch and integrates standard medical image processing libraries to efficiently process images during training of neural networks. TorchIO transforms can be easily composed, reproduced, traced and extended. Most transforms can be inverted, making the library suitable for test-time augmentation and estimation of aleatoric uncertainty in the context of segmentation. We provide multiple generic preprocessing and augmentation operations as well as simulation of MRI-specific artifacts.

RESULTS

Source code, comprehensive tutorials and extensive documentation for TorchIO can be found at http://torchio.rtfd.io/. The package can be installed from the Python Package Index (PyPI) running pip install torchio. It includes a command-line interface which allows users to apply transforms to image files without using Python. Additionally, we provide a graphical user interface within a TorchIO extension in 3D Slicer to visualize the effects of transforms.

CONCLUSION

TorchIO was developed to help researchers standardize medical image processing pipelines and allow them to focus on the deep learning experiments. It encourages good open-science practices, as it supports experiment reproducibility and is version-controlled so that the software can be cited precisely. Due to its modularity, the library is compatible with other frameworks for deep learning with medical images.

Hyperpolarized 129 Xe multi-slice imaging of the human brain using a 3D gradient echo pulse sequence

PURPOSE

To demonstrate the possibility of performing multi-slice in-vivo human brain MRI using hyperpolarized (HP) xenon-129 (¹²⁹ Xe) in two different orientations and to calculate the signal-to-noise ratio (SNR).

METHODS

Two healthy female participants were imaged during a single breath-hold of HP ¹²⁹ Xe using a Philips Achieva 3.0T MRI scanner (Philips, Andover, MA). Each HP ¹²⁹ Xe multi-slice brain image was acquired during separate HP ¹²⁹ Xe breath-holds using 3D gradient echo (GRE) imaging. The acquisition started 10 s after the inhalation of 1 L of HP ¹²⁹ Xe. Overall, four sagittal and three axial images were acquired (seven imaging sessions per participant). The SNR was calculated for each slice in both orientations.

RESULTS

The first ever HP ¹²⁹ Xe multi-slice images of the brain were acquired in axial and sagittal orientations. The HP ¹²⁹ Xe signal distribution correlated well with the gray matter distribution. The highest SNR values were close in the axial and sagittal orientations (19.46 ± 3.25 and 18.76 ± 4.94, respectively). Additionally, anatomical features, such as the ventricles, were observed in both orientations.

CONCLUSION

The possibility of using multi-slice HP ¹²⁹ Xe human brain magnetic resonance imaging was demonstrated for the first time. HP ¹²⁹ Xe multi-slice MRI can be implemented for brain imaging to improve current diagnostic methods.

MRI of Traumatic Knee Dislocation: A Study to Evaluate Safety and Image Quality for Patients with Knee-Spanning Stabilization Devices

This study evaluated safety and image quality of MRI exams performed for patients with traumatic knee dislocations in knee-spanning stabilization devices. It is an IRB-approved retrospective design with waived informed consent that included 63 patients with traumatic knee dislocation. 56 patients had metallic external fixators, and 7 patients had non-metallic knee immobilizers. 7 patients had bilateral dislocations yielding a total of 70 knee MRIs. 1.5 Tesla MRI exams were performed for all patients who were awake and alert at the time of imaging. All knee-spanning external fixators were considered "MR conditional" by the FDA. The electronic medical record was reviewed for notes from the technologist and nursing staff documenting any patient complaints or adverse events during the MRI exam as required by departmental protocol. Qualitative analysis of the six most frequently performed sequences were independently conducted by 2 musculoskeletal radiologists using a 5-point Likert scale. Overall image quality and select time intervals between the two groups were compared using an independent sample t test and the Mann-Whitney U test, respectively. No adverse events were reported for a 40-minute average estimated patient scan time with the stabilization devices in the MR gantry. Mean values of Likert scale scores were generated from two readers' data for comparison between the external fixation and the immobilizer groups. Most knee MRI exams with external fixators were within diagnostic quality despite artifacts (grade 3). MRI exams generally were of higher diagnostic quality in the immobilizer group than the external fixator group (p < 0.05). The external fixator models included DePuy Synthes, Smith and Nephew, Stryker Hoffman III, Zimmer FastFrame, and Zimmer XtraFix. MRI examinations in patients with external fixators for traumatic knee dislocations can be safely performed under certain conditions and provide diagnostic quality images.

Effect of breath holding at the end of the inspiration and expiration phases on liver stiffness measured by 2D-MR elastography

PURPOSE

Evaluate the effect of breath holding at the end of the inspiration and expiration phases on the measurement of liver stiffness by 2D-MR elastography (MRE).

METHODS

The study included 61 subjects, of which 31 subjects were pathologically confirmed liver fibrosis patients [liver fibrosis group (LFG)] and 30 healthy subjects [healthy group (HG)]. MRE was used to measure the liver stiffness of subjects in inspiration breath-hold (IBH) and expiration breath-hold (EBH) states.

RESULTS

In IBH and EBH states, the liver stiffness measured by MRE was not significantly different in the HG (P = 0.125, > 0.05), while the LFG showed a significant difference (P <0.001). Also, a significant difference was observed between the change values in the mild/moderate fibrosis group (MFG) and advanced fibrosis group (AFG) (P = 0.005, < 0.05).

CONCLUSIONS

The liver stiffness values in patients with liver fibrosis were affected by breath-holding states. The higher the stage of liver fibrosis, the greater the change in liver stiffness values, but no significant difference was observed between liver stiffness values in healthy subjects under the two breath-holding conditions. Different breath-holding states are factors influencing liver fibrosis stiffness measured by MRE, which should be given due attention in both clinical and research contexts.

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.

Forensic age estimation based on fast spin-echo proton density (FSE PD)-weighted MRI of the distal radial epiphysis

Radiation exposure is a crucial factor to consider in forensic age estimation. The various magnetic resonance imaging (MRI) modalities used in forensic age estimation avoid radiation exposure. This study examined the reliability of distal radius ossification using fast spin-echo proton density (FSE PD)–weighted MRI to estimate age. Left wrist MRI findings of 532 patients aged 10–29 years were evaluated retrospectively using the five-stage system of Dedouit et al. The intra- and interobserver reliability values were κ = 0.906 and 0.869, respectively. Based on the results, the respective minimum ages estimated for stages 4 and 5 were 13.4 and 16.1 years for females, and 15.1 and 17.3 years for males; the method could not estimate an age of 18 years in any case. FSE PD MRI analysis of the distal radius epiphysis provides supportive data and can be used when evaluating the distal radius for forensic age estimation.

Magnetic resonance microscopy of samples with translational symmetry with FOVs smaller than sample size

In MRI, usually the Field of View (FOV) has to cover the entire object. If this condition is not fulfilled, an infolding image artifact is observed, which suppresses visualization. In this study it is shown that for samples with translational symmetry, i.e., those consisting of identical objects in periodic unit cells, the FOV can be reduced to match the unit cell which enables imaging of an average object, of which the signal is originated from all unit cells of the sample, with no punishment by a loss in signal-to-noise ratio (SNR). This theoretical prediction was confirmed by experiments on a test sample with a 7 × 7 mm² unit cell arranged in a 3 × 3 matrix which was scanned by the spin-echo and by single point imaging methods. Effects of experimental imperfections in size and orientation mismatch between FOV and unit cell were studied as well. Finally, this method was demonstrated on a 3D periodic sample of tablets, which yielded well-resolved images of moisture distribution in an average tablet, while single tablet imaging provided no results. The method can be applied for SNR increase in imaging of any objects with inherently low signals provided they can be arranged in a periodic structure.

Imaging β-Cell Function Using a Zinc-Responsive MRI Contrast Agent May Identify First Responder Islets

An imaging method for detecting β-cell function in real-time in the rodent pancreas could provide new insights into the biological mechanisms involving loss of β-cell function during development of type 2 diabetes and for testing of new drugs designed to modulate insulin secretion. In this study, we used a zinc-responsive MRI contrast agent and an optimized 2D MRI method to show that glucose stimulated insulin and zinc secretion can be detected as functionally active "hot spots" in the tail of the rat pancreas. A comparison of functional images with histological markers show that insulin and zinc secretion does not occur uniformly among all pancreatic islets but rather that some islets respond rapidly to an increase in glucose while others remain silent. Zinc and insulin secretion was shown to be altered in streptozotocin and exenatide treated rats thereby verifying that this simple MRI technique is responsive to changes in β-cell function.

Comparing the clinical utility of single-shot echo-planar imaging and readout-segmented echo-planar imaging in diffusion-weighted imaging of the liver at 3 tesla

PURPOSE

To compare the clinical utility of single-shot echo-planar imaging (SS-EPI) using different breathing schemes and readout-segmented EPI (RS-EPI) in the repeatability of apparent diffusion coefficient (ADC) measurements, signal-to-noise ratio (SNR) and image quality.

METHODS

In this institutional review board-approved prospective study, hepatic DWIs (b = 50, 300, 600 s/mm²) were performed in 22 volunteers on 3.0 T MRI using SS-EPI with free-breathing diffusion-weighted imaging (FB-DWI), breath-hold (BH-DWI), respiratory-triggered (RT-DWI) and navigator-triggered (NT-DWI), and readout-segmented EPI (RS-DWI). ADC and surrogate SNR (sSNR) were measured in nine anatomic locations in the right lobe, and image quality was assessed on all FB-DWI, BH-DWI, RT-DWI, NT-DWI, and RS-DWI sequences. The sequence with the optimal clinical utility was decided by systematically comparing the ADC repeatability, sSNR and image quality of the above DWIs.

RESULTS

In all the five sequences, NT-DWI had the most reliable intra-observer agreement (intraclass correlation coefficient (ICC): 0.900-0.922; all P > 0.05), and a better interobserver agreement (ICC: 0.853-0.960; all p > 0.05) than RS-DWI (ICC:0.881-0.916; some P < 0.05). NT-DWI had the best ADC repeatability in the nine locations (mean ADC absolute differences: 38.47-56.38 × 10^-6 mm²/s, limits of agreement (LOA): 17.33-22.52 × 10^-6 mm²/s). Also, NT-DWI had the highest sSNR (Reader 1: 50.58 ± 20.11 (Superior), 74.06 ± 28.37 (Central), 80.99 ± 38.11(Inferior)); Reader 2: 48.07 ± 23.92 (Superior), 68.23 ± 32.91 (Central), 76.78 ± 33.07 (Inferior)) in three representative sections except for RS-DWI. Furthermore, NT-DWI had a better image quality than RS-DWI (P < 0.05) and was superior to FB-DWI and BH-DWI in sharpness of the liver (at b = 300 s/mm²) (P < 0.05) CONCLUSION: RS-DWI has the best SNR. However, NT-DWI can provide sufficient SNR, excellent image quality, and the best ADC repeatability on 3.0 T MRI. It is thus the recommended sequence for the clinical application of hepatic DWI.

MR safety considerations for patients undergoing prostate MRI

Over the past decade, there has been a dramatic increase in the number of patients undergoing prostate MRI scans. Patients presenting for prostate MRI are an ageing population and may present with a variety of passive or active implants and devices. These implants and devices can be MR safe or MR conditional or MR unsafe. Patients with certain MR-conditional active implants and devices can safely obtain prostate MRI in a specified MR environment within specific MR imaging parameters. Prostate MRI and PET-MRI in patients with passive implants such as hip prostheses, fiducial markers for SBRT, brachytherapy seeds and prostatectomy bed clips have unique concerns for image optimization that can cause geometric distortion of the diffusion-weighted imaging (DWI) sequence. We discuss strategies to overcome these susceptibility artifacts. Prostate MRI in patients with MR conditional active implants such as cardiac implantable electronic devices (CIED) also require modification of imaging parameters and magnet strength. In this setting, a diagnostic quality prostate MRI can be performed at a lower magnet strength (1.5 T) along with modification of imaging parameters to ensure patient safety. Imaging strategies to minimize susceptibility artifact and decrease the specific absorption rate (SAR) in both settings are described. Knowledge of MR safety considerations and imaging strategies specific to prostate MRI and PET-MRI in patients with implants and devices is essential to ensure diagnostic-quality MR images and patient safety.

Characterization of magnetic interference and image artefacts during simultaneous in-beam MR imaging and proton pencil beam scanning

For the first time, a low-field open magnetic resonance (MR) scanner was combined with a proton pencil beam scanning (PBS) research beamline. The aim of this study was to characterize the magnetic fringe fields produced by the PBS system and measure their effects on MR image quality during simultaneous PBS irradiation and image acquisition. A magnetic field camera measured the change in central resonance frequency (Δf _res) and magnetic field homogeneity (ΔMFH) of the B₀ field of the MR scanner during operation of the beam transport and scanning magnets. The beam energy was varied between 70 - 220 MeV and beam scanning was performed along the central horizontal and vertical axis of a 48 × 24 cm² radiation field. The time structure of the scanning magnets' fringe fields was simultaneously recorded by a tri-axial Hall probe. MR imaging experiments were conducted using the ACR (American College of Radiology) Small MRI Phantom and a spoiled gradient echo pulse sequence during simultaneous volumetric irradiation. Computer simulations were performed to predict the effects of B ₀ field perturbations due to PBS irradiation on MR image formation in k-space. Setting the beam transport magnets, horizontal and vertical scanning magnets resulted in a maximum Δf _res of 50, 235 and 4 Hz, respectively. The ΔMFH was less than 3 parts per million for all measurements. MR images acquired during beam energy variation and vertical beam scanning showed no visual loss in image quality. However, MR images acquired during horizontal beam scanning showed severe coherent ghosting artefacts in phase encoding direction. Both simulated and measured k-space phase maps prove that these artefacts are caused by phase-offsets. This study shows first experimental evidence that simultaneous in-beam MR imaging during proton PBS irradiation is subject to severe loss of image quality in the absence of magnetic decoupling between the PBS and MR system.

Evaluation of simultaneous-multislice diffusion-weighted imaging of liver at 3.0 T with different breathing schemes

PURPOSE

To obtain the optimal simultaneous-multislice (SMS)-accelerated diffusion-weighted imaging (DWI) of the liver at 3.0 T MRI by systematically estimating the repeatability of apparent diffusion coefficient (ADC), signal-to-noise ratio (SNR) and image quality of different breathing schemes in comparison to standard DWI (STD) and other SMS sequences.

METHODS

In this institutional review board-approved prospective study, hepatic DWIs (b = 50, 300, 600 s/mm²) were performed in 23 volunteers on 3.0 T MRI using SMS and STD with breath-hold (BH-SMS, BH-STD), free-breathing (FB-SMS, FB-STD) and respiratory-triggered (RT-SMS, RT-STD). Reduction of scan time with SMS-acceleration was calculated. ADC and SNR were measured in nine anatomic locations and image quality was assessed on all SMS and STD sequences. An optimal SMS-DWI was decided by systematically comparing the ADC repeatability, SNR and image quality among above DWIs.

RESULTS

SMS-DWI reduced scan time significantly by comparison with corresponding STD-DWI (27 vs. 42 s for BH, 54 vs. 78 s for FB and 42 vs. 97 s for RT). In all DWIs, BH-SMS had the greatest intraobserver agreement (intraclass correlation coefficient (ICC): 0.920-0.944) and good interobserver agreement (ICC: 0.831-0.886) for ADC measurements, and had the best ADC repeatability (mean ADC absolute differences: 0.046-0.058 × 10^-3mm²/s, limits of agreement (LOA): 0.010-0.013 × 10^-3mm²/s) in nine locations. BH-SMS had the highest SNR in three representative sections except for RT-STD. There were no significant differences in image quality between BH-SMS and other DWI sequences (median BH-SMS: 4.75, other DWI: 4.5-5.0; P > 0.0.5).

CONCLUSION

BH-SMS provides considerable scan time reduction with good image quality, sufficient SNR and highest ADC repeatability on 3.0 T MRI, which is thus recommended as the optimal hepatic DWI sequence for those subjects with adequate breath-holding capability.

A modern review of the uncertainties in volumetric imaging of respiratory-induced target motion in lung radiotherapy

Radiotherapy has become a critical component for the treatment of all stages and types of lung cancer, often times being the primary gateway to a cure. However, given that radiation can cause harmful side effects depending on how much surrounding healthy tissue is exposed, treatment of the lung can be particularly challenging due to the presence of moving targets. Careful implementation of every step in the radiotherapy process is absolutely integral for attaining optimal clinical outcomes. With the advent and now widespread use of stereotactic body radiation therapy (SBRT), where extremely large doses are delivered, accurate, and precise dose targeting is especially vital to achieve an optimal risk to benefit ratio. This has largely become possible due to the rapid development of image-guided technology. Although imaging is critical to the success of radiotherapy, it can often be plagued with uncertainties due to respiratory-induced target motion. There has and continues to be an immense research effort aimed at acknowledging and addressing these uncertainties to further our abilities to more precisely target radiation treatment. Thus, the goal of this article is to provide a detailed review of the prevailing uncertainties that remain to be investigated across the different imaging modalities, as well as to highlight the more modern solutions to imaging motion and their role in addressing the current challenges.

Description of the MRI and ultrasound imaging features of giant epidermal cysts

OBJECTIVES

Guidelines suggest that lesions over 5 cm in dimension should be referred to a specialist sarcoma centre due to the possibility of malignancy. Few epidermal cysts (ECs) reach or exceed this size and are termed giant ECs (GECs). The purpose of this study is to report on a large series of GECs.

METHODS

Retrospective review of histologically proven GECs over an 8-year period. Patient demographics with MRI and ultrasound (US) appearances were evaluated.

RESULTS

A total of 14 cases were included with eight males and six females. Mean age was 51 years. 11 lesions were oval and three bi-lobed in shape, while 12 demonstrated dermal apposition. All were hyperintense on water-sensitive sequences and isointense to slightly hypointense on T1W imaging. Internal clefts were seen in 13 cases and 11 demonstrated chemical shift artefact (CSA) on MRI. On US, 12 showed well-defined linear hypoechoic clefts, with 66.6% having dis-organised compared with 33.3% peripherally located clefts. One 'pseudo testis' pattern and one showing irregular striped echogenicity termed novel 'pseudo muscle' appearance. No cases demonstrated internal vascularity on Doppler US.

CONCLUSIONS

MRI signal findings of GECs are often characteristic with hyperintensity on water-sensitive sequences, dermal apposition, CSA and internal clefts while US features of disorganised or clumped hypoechoic clefts and absence of neovascularity were commonly seen. Recognition of combinations of both US and MRI features of GECs should reduce the requirement for pre-excisional needle biopsy to confirm the diagnosis.

ADVANCES IN KNOWLEDGE

1. Identification of common imaging features of GECs should avoid unnecessary pre-excisional biopsy despite their large size in the appropriate MDT setting.2. A novel 'pseudo-muscle' appearance is described on MRI and US.

Assessment of cardiac function, blood flow and myocardial tissue relaxation parameters at 0.35 T

A low field strength (B0) system could increase cardiac MRI availability for patients otherwise contraindicated at higher field. Lower equipment costs could also broaden cardiac MR accessibility. The current study investigated the feasibility of cardiac function with steady-state free precession and flow assessment with phase contrast (PC) cine images at 0.35 T, and evaluated differences in myocardial relaxation times using quantitative T1, T2 and T2* maps by comparison with 1.5 and 3 T results in a small cohort of six healthy volunteers. Signal-to-noise ratio (SNR) differences across systems were characterized with proton density-weighted spin echo phantom data. SNR at 0.35 T was lower by factors of 5.5 and 15.0 compared with the 1.5 and 3 T systems used in this study. All cine images at 0.35 T scored 3 or greater on a five-point image quality scale. Normalized blood-myocardium contrast in cine images, left ventricular volumes (end diastolic volume, end systolic volume) and function (ejection fraction and stroke volume) measures at 0.35 T matched 1.5 and 3 T results. Phase-to-noise ratio in 0.35 T PC images (11.7 ± 1.9) was lower than 1.5 T (18.7 ± 5.2) and 3 T (44.9 ± 16.5). Peak velocity and stroke volume determined from PC images were similar across systems. Myocardial T1 increased (564 ± 13 ms at 0.35 T, 955 ± 19 ms at 1.5 T and 1200 ± 35 ms at 3 T) while T2 (59 ± 4 ms at 0.35 T, 49 ± 3 ms at 1.5 T and 40 ± 2 ms at 3 T) and T2* (42 ± 8 ms at 0.35 T, 33 ± 6 ms at 1.5 T and 24 ± 3 ms at 3 T) decreased with increasing B0. Despite SNR deficits, cardiovascular function, flow assessment and myocardial relaxation parameter mapping is feasible at 0.35 T using standard cardiovascular imaging sequences.

Prostate magnetic resonance imaging technique

Multiparametric magnetic resonance (MR) imaging of the prostate is an excellent tool to detect clinically significant prostate cancer, and it has widely been incorporated into clinical practice due to its excellent tissue contrast and image resolution. The aims of this article are to describe the prostate MR imaging technique for detection of clinically significant prostate cancer according to PI-RADS v2.1, as well as alternative sequences and basic aspects of patient preparation and MR imaging artifact avoidance.

MRI restoration using edge-guided adversarial learning

Magnetic resonance imaging (MRI) images acquired as multislice two-dimensional (2D) images present challenges when reformatted in orthogonal planes due to sparser sampling in the through-plane direction. Restoring the "missing" through-plane slices, or regions of an MRI image damaged by acquisition artifacts can be modeled as an image imputation task. In this work, we consider the damaged image data or missing through-plane slices as image masks and proposed an edge-guided generative adversarial network to restore brain MRI images. Inspired by the procedure of image inpainting, our proposed method decouples image repair into two stages: edge connection and contrast completion, both of which used general adversarial networks (GAN). We trained and tested on a dataset from the Human Connectome Project to test the application of our method for thick slice imputation, while we tested the artifact correction on clinical data and simulated datasets. Our Edge-Guided GAN had superior PSNR, SSIM, conspicuity and signal texture compared to traditional imputation tools, the Context Encoder and the Densely Connected Super Resolution Network with GAN (DCSRN-GAN). The proposed network may improve utilization of clinical 2D scans for 3D atlas generation and big-data comparative studies of brain morphometry.

Real-Time Quality Assessment of Pediatric MRI via Semi-Supervised Deep Nonlocal Residual Neural Networks

In this paper, we introduce an image quality assessment (IQA) method for pediatric T1- and T2-weighted MR images. IQA is first performed slice-wise using a nonlocal residual neural network (NR-Net) and then volume-wise by agglomerating the slice QA results using random forest. Our method requires only a small amount of quality-annotated images for training and is designed to be robust to annotation noise that might occur due to rater errors and the inevitable mix of good and bad slices in an image volume. Using a small set of quality-assessed images, we pre-train NR-Net to annotate each image slice with an initial quality rating (i.e., pass, questionable, fail), which we then refine by semi-supervised learning and iterative self-training. Experimental results demonstrate that our method, trained using only samples of modest size, exhibit great generalizability, capable of real-time (milliseconds per volume) large-scale IQA with nearperfect accuracy.

Large-scale Quality Control of Cardiac Imaging in Population Studies: Application to UK Biobank

In large population studies such as the UK Biobank (UKBB), quality control of the acquired images by visual assessment is unfeasible. In this paper, we apply a recently developed fully-automated quality control pipeline for cardiac MR (CMR) images to the first 19,265 short-axis (SA) cine stacks from the UKBB. We present the results for the three estimated quality metrics (heart coverage, inter-slice motion and image contrast in the cardiac region) as well as their potential associations with factors including acquisition details and subject-related phenotypes. Up to 14.2% of the analysed SA stacks had sub-optimal coverage (i.e. missing basal and/or apical slices), however most of them were limited to the first year of acquisition. Up to 16% of the stacks were affected by noticeable inter-slice motion (i.e. average inter-slice misalignment greater than 3.4 mm). Inter-slice motion was positively correlated with weight and body surface area. Only 2.1% of the stacks had an average end-diastolic cardiac image contrast below 30% of the dynamic range. These findings will be highly valuable for both the scientists involved in UKBB CMR acquisition and for the ones who use the dataset for research purposes.

Identification of Naturally Occurring Cartilage Damage in the Equine Distal Interphalangeal Joint Using Low-Field Magnetic Resonance Imaging and Magnetic Resonance Arthrography

Objectives: To describe detectable and non-detectable naturally occurring cartilage damage of the equine distal interphalangeal (DIP) joint using plain magnetic resonance (MR) imaging and gadolinium and saline MR arthrography. The second objective was to quantify the sensitivity, specificity and accuracy in detection of cartilage damage.

Methods: In a pilot study, the distal limbs of two horses with confirmed osteoarthritis of the DIP joint were imaged with low-field MR. Magnetic resonance images were assessed in consensus by three observers and compared to gross pathological findings. Subsequently, a prospective analytical cross-sectional study design was created to compare pre-contrast MR imaging and saline and gadolinium MR arthrography of isolated equine distal limbs to gross observation findings. Hallmarq® low-field MR (0.27T) scans were performed prior to DIP joint injection, saline/gadolinium post-injection scans were performed at 15 min intervals for 2 h. Joints were inspected and the articular cartilage graded subjectively for cartilage damage (0–3). The presence of detectable or non-detectable cartilage damage on MR images was identified, characterized and recorded in consensus by three observers. Sensitivity, specificity and accuracy in detection of cartilage damage related to gross pathology were calculated.

Results: The two clinical cases from the pilot study with confirmed osteoarthritis had full thickness cartilage defects; however, only one of these was correctly identified using low-field MRI. In the prospective study, the majority of naturally occurring cartilage damage could not be identified on plain MR or MR arthrography including extensive partial thickness cartilage erosions. Saline and gadolinium MR arthrography did not improve the detection of cartilage damage. The accuracy of cartilage damage detection was 0.63 with a sensitivity of 0.14 and specificity of 0.92.

Clinical Relevance: Both, plain low-field MRI and MR arthrography are not sensitive in detection of naturally occurring cartilage damage of the DIP joint. However, if an abnormal contour is seen in the articular cartilage, cartilage damage is likely to be present.

MRI-Based Upper Abdominal Organs-at-Risk Atlas for Radiation Oncology

PURPOSE

The purpose of our study was to provide a guide for identification and contouring of upper abdominal organs-at-risk (OARs) in the setting of online magnetic resonance imaging (MRI)-guided radiation treatment planning and delivery.

METHODS AND MATERIALS

After a needs assessment survey, it was determined that an upper abdominal MRI-based atlas of normal OARs would be of benefit to radiation oncologists and radiation therapists. An anonymized diagnostic 1.5T MRI from a patient with typical upper abdominal anatomy was used for atlas development. Two MRI sequences were selected for contouring, a T1-weighted gadoxetic acid contrast-enhanced MRI acquired in the hepatobiliary phase and axial fast imaging with balanced steady-state precession. Two additional clinical MRI sequences from commercial online MRI-guided radiation therapy systems were selected for contouring and were included in the final atlas. Contours from each data set were completed and reviewed by radiation oncologists, along with a radiologist who specializes in upper abdominal imaging, to generate a consensus upper abdominal MRI-based OAR atlas.

RESULTS

A normal OAR atlas was developed, including recommendations for contouring. The atlas and contouring guidance are described, and high-resolution MRI images and contours are displayed. OARs, such as the bile duct and biliary tree, which may be better seen on MRI than on computed tomography, are highlighted. The full DICOM/DICOM-RT MRI images from both the diagnostic and clinical online MRI-guided radiation therapy systems data sets have been made freely available, for educational purposes, at econtour.org.

CONCLUSIONS

This MRI contouring atlas for upper abdominal OARs should provide a useful reference for contouring and education. Its routine use may help to improve uniformity in contouring in radiation oncology planning and OAR dose calculation. Full DICOM/DICOM-RT images are available online and provide a valuable educational resource for upper abdominal MRI-based radiation therapy planning and delivery.

Comparison of magnetization transfer contrast of conventional and simultaneous multislice turbo spin echo acquisitions focusing on excitation time interval

PURPOSE

Image contrast differs between conventional multislice turbo spin echo (conventional TSE) and multiband turbo spin echo (SMS-TSE). Difference in time interval between excitations for adjacent slices (SETI) might cause this difference. This study aimed to evaluate the influence of SETI on MT effect for conventional TSE and compare conventional TSE with SMS-TSE in this respect.

MATERIALS AND METHODS

Three different agar concentration phantoms were scanned with conventional TSE by adjusting SETI and TR. Signal change for different SETI was evaluated using Pearson's correlation analysis. SMS-TSE was acquired by changing TR similarly. Three human volunteers were scanned with similar settings to evaluate reproducibility of the phantom results in human brain.

RESULTS

In conventional TSE, shorter SETI induced larger signal reduction. Longer TR and higher agar concentration emphasized this characteristic. Significant linear correlation (P < 0.05) was found in the major cases. The SMS-TSE signal intensity in each TR and phantom was smaller than the assumable levels in conventional TSE when the slices were simultaneously excited. Similar characteristic was observed in human brain.

CONCLUSION

Shorter SETI results in larger MT effect in conventional TSE. The contrast change in SMS-TSE was larger than the supposable level from simultaneous excitation, which needs consideration in clinics.

Learning-Based Quality Control for Cardiac MR Images

The effectiveness of a cardiovascular magnetic resonance (CMR) scan depends on the ability of the operator to correctly tune the acquisition parameters to the subject being scanned and on the potential occurrence of imaging artifacts, such as cardiac and respiratory motion. In the clinical practice, a quality control step is performed by visual assessment of the acquired images; however, this procedure is strongly operator-dependent, cumbersome, and sometimes incompatible with the time constraints in clinical settings and large-scale studies. We propose a fast, fully automated, and learning-based quality control pipeline for CMR images, specifically for short-axis image stacks. Our pipeline performs three important quality checks: 1) heart coverage estimation; 2) inter-slice motion detection; 3) image contrast estimation in the cardiac region. The pipeline uses a hybrid decision forest method-integrating both regression and structured classification models-to extract landmarks and probabilistic segmentation maps from both long- and short-axis images as a basis to perform the quality checks. The technique was tested on up to 3000 cases from the UK Biobank and on 100 cases from the UK Digital Heart Project and validated against manual annotations and visual inspections performed by expert interpreters. The results show the capability of the proposed pipeline to correctly detect incomplete or corrupted scans (e.g., on UK Biobank, sensitivity and specificity, respectively, 88% and 99% for heart coverage estimation and 85% and 95% for motion detection), allowing their exclusion from the analyzed dataset or the triggering of a new acquisition.

A deep learning framework for efficient analysis of breast volume and fibroglandular tissue using MR data with strong artifacts

PURPOSE

The main purpose of this work is to develop, apply, and evaluate an efficient approach for breast density estimation in magnetic resonance imaging data, which contain strong artifacts including intensity inhomogeneities.

METHODS

We present a pipeline for breast density estimation, which consists of intensity inhomogeneity correction, breast volume segmentation, nipple extraction, and fibroglandular tissue segmentation. For the segmentation steps, a well-known deep learning architecture is employed.

RESULTS

The average Dice coefficient for the breast parenchyma is [Formula: see text], which outperforms the classical state-of-the-art approach by a margin of [Formula: see text].

CONCLUSION

The proposed solution is accurate and highly efficient and has potential to be applied for big epidemiological data with thousands of participants.