3D Magnetic Resonance Imaging Research Articles

Hypoperfusion of hypothalamic subunits in medication-overuse headache using a 3D PCASL MRI

Hypothalamus is a crucial deep brain area that is responsible for the integration and coordination of various brain functions. The altered perfusion of hypothalamus during headache caused by medication-overuse headache (MOH) was previously unknown. In the current study, the altered perfusion of hypothalamic subregions in MOH patients was investigated using state-of-the-art 3D pseudo-continuous arterial spin labeling (PCASL) MR imaging. In this study, 29 normal controls subjects (NCs) and 29 MOH patients underwent 3D PCASL and brain structural MR imaging. The hypothalamus was automatically segmented into 10 subunits and the volume of each subunit was automatically determined using Freesurfer software (v7.4.1). All segmented hypothalamic subunits were converted to individual hypothalamic subunit masks. The cerebral blood flow (CBF) images were coregistered with the raw brain structural images and resliced. The CBF value of each hypothalamic subunit was extracted from the warped CBF images. The volume and CBF value of each hypothalamic subunit were analyzed using the independent sample T test and Mann–Whitney U test, receiver operating characteristic (ROC) curve analysis, and Pearson and Spearman correlation analysis. Hypothalamic subunits with significantly decreased perfusion were located in the left posterior, left tubular superior, right anterior–inferior, right tubular inferior, right tubular superior, right posterior subunit and the entire right hypothalamus [CBF value for MOH vs NC (mL/100 g·min): 48.41 ± 6.75 vs 54.08 ± 11.47, 44.44 ± 4.79 vs 48.11 ± 7.73, 41.49 (32.90, 61.46) vs 49.38 ± 10.47, 46.62 ± 7.04 vs 53.90 ± 11.75, 42.12 ± 5.74 vs 47.02 ± 9.99, 42.79 ± 5.15 vs 47.93 ± 10.48 and 43.58 ± 5.06 vs 48.65 ± 9.33, respectively] in MOH compared to NC (P < 0.05). ROC analysis for these positive subunits revealed that area under the curve was 0.658–0.693, and ROC curve for left posterior subunit had the highest specificity of 93.10% while the entire right hypothalamus had the highest sensitivity of 72.41%. Further correlation analysis showed that the CBF value of the left posterior, right anterior–inferior, right tubular superior, whole right hypothalamus presented significantly negative correlation with Hamilton Depression Scale (HAMD) score (P < 0.05). Hypoperfusion of hypothalamic subunits may contribute to the understanding of MOH pathogenesis, and the 3D PCASL could be considered as a potential diagnostic and assessment tool for MOH.

NCNet: deformable medical image registration network based on neighborhood cross-attention combined with multi-resolution constraints.

&#xD;Existing registration networks based on cross-attention design usually divide the image pairs to be registered into patches for input. The division and merging operations of a series of patches are difficult to maintain the topology of the deformation field and reduce the interpretability of the network. Therefore, our goal is to develop a new network architecture based on a cross-attention mechanism combined with a multi-resolution strategy to improve the accuracy and interpretability of medical image registration.&#xD;Approach: We propose a new deformable image registration network NCNet based on neighborhood cross-attention combined with multi-resolution strategy. The network structure mainly consists of a multi-resolution feature encoder, a multi-head neighborhood cross-attention module and a registration decoder. The hierarchical feature extraction capability of our encoder is improved by introducing large kernel parallel convolution blocks; the cross-attention module based on neighborhood calculation is used to reduce the impact on the topology of the deformation field and double normalization is used to reduce its computational complexity.&#xD;Main result: We performed atlas-based registration and inter-subject registration tasks on the public 3D brain magnetic resonance imaging datasets LPBA40 and IXI respectively. Compared with the popular VoxelMorph method, our method improves the average DSC value by 7.9% and 3.6% on LPBA40 and IXI. Compared with the popular TransMorph method, our method improves the average DSC value by 4.9% and 1.3% on LPBA40 and IXI.&#xD;Significance: We proved the advantages of the neighborhood attention calculation method compared to the window attention calculation method based on partitioning patches, and analyzed the impact of the pyramid feature encoder and double normalization on network performance. This has made a valuable contribution to promoting the further development of medical image registration methods.&#xD.

Dementia with Lewy bodies and gait neural basis: a cross-sectional study

BackgroundDementia with Lewy Bodies (DLB) is responsible for cognitive-behavioural disorders but also for gait disorders. The latter are thought to be related to parkinsonism, but the neural bases of these disorders are not well known, especially in the early stages. The aim of this study was to investigate by volumetric Magnetic Resonance Imaging the neuronal basis of gait disorders in DLB patients, compared to Healthy Elderly Controls and Alzheimer’s Disease patients.MethodsClinical examination with motor assessment including 10-meter walking speed, one-leg balance and Timed Up and Go test, a comprehensive neuropsychological evaluation and 3D brain Magnetic Resonance Imaging were performed on 84 DLB patients, 39 Alzheimer’s Disease patients and 22 Healthy Elderly Controls. We used Statistical Parametric Mapping 12 to perform a one-sample t-test to investigate the correlation between each gait score and gray matter volume (P ≤ 0.05 corrected for family-wise error).ResultsWe found a correlation for DLB patients between walking speed and gray matter decrease (P < 0.05, corrected for family-wise error) in caudate nuclei, anterior cingulate cortex, mid-cingulate cortex, hippocampi, supplementary motor area, right cerebellar cortex and left parietal operculum. We found no correlation with Timed Up and Go test and one-leg balance.ConclusionGait disorders are underpinned by certain classical regions such as the cerebellum and the supplementary motor area. Our results suggest there may be a motivational and emotional component of voluntary gait in DLB subjects, underpinned by the cingulate cortex, a spatial orientation component, underpinned by hippocampi and suggest the involvement of brain processing speed and parkinsonism, underpinned by the caudate nuclei.Trial registrationThe study protocol has been registered on ClinicalTrials.gov. (NCT01876459) on June 12, 2013.

Gender variability in machine learning based subcortical neuroimaging for Parkinson’s disease diagnosis

PurposeThis study evaluates machine learning (ML) classifiers for diagnosing Parkinson’s disease (PD) using subcortical brain region data from 3D T1 magnetic resonance imaging (MRI) Parkinson’s Progression Markers Initiative (PPMI database). We aim to identify top-performing algorithms and assess gender-related differences in accuracy.Design/methodology/approachMultiple ML algorithms will be compared for their ability to classify PD vs healthy controls using MRI scans of the brain structures like the putamen, thalamus, brainstem, accumbens, amygdala, caudate, hippocampus and pallidum. Analysis will include gender-specific performance comparisons.FindingsThe study reveals that ML classifier performance in diagnosing PD varies across subcortical brain regions and shows gender differences. The Extra Trees classifier performed best in men (86.36% accuracy in the putamen), while Naive Bayes performed best in women (69.23%, amygdala). Regions like the accumbens, hippocampus and caudate showed moderate accuracy (65–70%) in men and poor performance in women. The results point out a significant gender-based performance gap, highlighting the need for gender-specific models to improve diagnostic precision across complex brain structures.Originality/valueThis study highlights the significant impact of gender on machine learning diagnosis of PD using data from subcortical brain regions. Our novel focus on these regions uncovers their diagnostic potential, improves model accuracy and emphasizes the need for gender-specific approaches in medical AI. This work could ultimately lead to earlier PD detection and more personalized treatment.

Assessment of 4D flow MRI for quantification of left-to-right shunt in pediatric patients with ventricular septal defect: comparison with right heart catheterization.

The percentage of shunt fraction significantly impacts the management of patients with congenital shunts, influencing strategic choices such as surgical or interventional procedures. This study compared the estimated shunt fraction (the ratio of pulmonary-to-systemic flow, Qp/Qs) for quantifying the left-to-right shunt in children with ventricular septal defect (VSD) using heart catheterization, four-dimensional (4D) flow, and two-dimensional (2D) flow magnetic resonance imaging (MRI). The goal was to establish a non-invasive and reliable measurement ratio between pulmonary and systemic blood flow in these patients. Between July 2022 and June 2023, patients scheduled to undergo invasive right heart catheterization were included in this study. MRI was performed one hour before the catheterization procedure. The correlation of shunt fraction was assessed between all methods after calculating the Qp/Qs ratio from 2D and 4D flow MRI and catheterization. A total of 24 patients (aged 3-15 years, eight females) were ultimately included in the study. The Qp/Qs ratios obtained from 4D flow had a robust correlation (correlation coefficient r = 0.962) compared to those obtained during catheterization. Cardiac catheterization recorded the mean shunt fraction at 1.499 ± 0.396, while 4D flow measured it at 1.403 ± 0.344, with no significant difference between the two techniques. Moreover, there was a reasonable correlation (r = 0.894) between 2D flow measurements of Qp/Qs and the results obtained from catheterization, with a mean shunt fraction of 1.326 ± 0.283. 4D flow MRI has the potential to be a non-invasive method for accurately measuring the left-to-right shunt in children with VSD.

Comparison of Left Ventricular Function Derived from Subject-Specific Inverse Finite Element Modeling Based on 3D ECHO and Magnetic Resonance Images.

Three-dimensional echocardiography (3D ECHO) and magnetic resonance (MR) imaging are frequently used in patients and animals to evaluate heart functions. Inverse finite element (FE) modeling is increasingly applied to MR images to quantify left ventricular (LV) function and estimate myocardial contractility and other cardiac biomarkers. It remains unclear, however, as to whether myocardial contractility derived from the inverse FE model based on 3D ECHO images is comparable to that derived from MR images. To address this issue, we developed a subject-specific inverse FE model based on 3D ECHO and MR images acquired from seven healthy swine models to investigate if there are differences in myocardial contractility and LV geometrical features derived using these two imaging modalities. We showed that end-systolic and end-diastolic volumes derived from 3D ECHO images are comparable to those derived from MR images (R2=0.805 and 0.969, respectively). As a result, ejection fraction from 3D ECHO and MR images are linearly correlated (R2=0.977) with the limit of agreement (LOA) ranging from -17.95% to 45.89%. Using an inverse FE modeling to fit pressure and volume waveforms in subject-specific LV geometry reconstructed from 3D ECHO and MR images, we found that myocardial contractility derived from these two imaging modalities are linearly correlated with an R2 value of 0.989, a gradient of 0.895, and LOA ranging from -6.11% to 36.66%. This finding supports using 3D ECHO images in image-based inverse FE modeling to estimate myocardial contractility.

Computational fluid dynamics of bladder voiding using 3D dynamic MRI.

Over the last couple of decades, image-based computational fluid dynamics (CFD) has revolutionized cardiovascular research by uncovering hidden features of wall strain, impact of vortices, and its use in treatment planning, as examples, that were simply not evident in the gold-standard catheterization studies done previously. In the work presented here, we have applied magnetic resonance imaging (MRI)-based CFD to study bladder voiding and to demonstrate the feasibility and potential of this approach. We used 3D dynamic MRI to image the bladder and urethra during voiding. A surface mesh processing tool was developed to process the bladder wall prior to executing a wall-motion driven CFD simulation of the bladder and urethra. The obtained flow rate and pressure were used to calculate urodynamic nomograms, which are currently used in the clinical setting to assess bladder voiding dysfunction. These nomograms concluded that our healthy volunteer has an unobstructed bladder and normal contractility. We calculated the work done to void the bladder and propose this as an additional quantitative metric to comprehensively assess bladder function. Further, we discuss the areas that would improve this relatively new methodology of image-based CFD in urodynamics.

Enzyme replacement therapy for CLN2 disease: MRI volumetry shows significantly slower volume loss compared to a natural history cohort.

Neuronal ceroid lipofuscinoses (NCL) are a group of neurodegenerative disorders. Recently, enzyme replacement therapy (ERT) was approved for CLN2, a subtype of NCL. The aim of this study was to quantify brain volume loss in CLN2 disease of patients on ERT in comparison to a natural history cohort using magnetic resonance imaging (MRI). Nineteen (13 female, 6 male) patients with CLN2 disease at one UK center were studied using serial 3D T1-weighted MRI (follow-up time, 1 to 9 years). Brain segmentation was done using FreeSurfer. Volume measurements for supratentorial grey and white matter, deep grey matter (basal ganglia/thalami), lateral ventricles, and cerebellar grey and white matter were recorded. The volume change over time was analyzed using a linear mixed-effects model excluding scans before treatment start. Comparison was made to a published natural history cohort of 12 patients (8 female, 4 male) which was reanalyzed using the same method. Brain volume loss of all segmented brain regions was much slower in treated patients compared to the natural history cohort. For example, supratentorial grey matter volume in treated patients decreased by 3±0.74% (p<0.001) annually compared to an annual volume loss of 16.8±1.5% (p<0.001) in the natural history cohort. Our treatment cohort showed a significantly slower rate of brain parenchymal volume loss compared to a natural history cohort in several anatomical regions. Our results complement prior clinical data which found a positive response to ERT. We demonstrate that automated MRI volumetry is a sensitive tool to monitor treatment response in children with CLN2 disease. NCL = Neuronal Ceroid Lipofuscinosis, CLN2 = Neuronal Ceroid Lipofuscinosis type 2, TPP1 = tripeptidyl peptidase 1, ERT = enzyme replacement therapy, EMA = European Medicines Agency, ICV = intra-cerebro-ventricular reservoir.

The deltoid ligament complex is reliably visualised as four fascicles in two layers with 3D volumetric MRI in the ankles of 20 asymptomatic volunteers.

The anatomy of the deltoid ligament is complex. There is agreement on the presence of superficial and deep layers but the number and frequency of fascicles remains controversial. Identifying injuries to specific components of the deltoid ligament may inform decision-making on their management. The anatomy was reviewed to establish the number and dimension of fascicles visible with three-dimensional (3D) volumetric magnetic resonance images (MRI). Twenty ankles from asymptomatic healthy volunteers were imaged with 3D volumetric MRI. The presence of individual fascicles was recorded and measured in 3D. The median age of participants was 26 years (range: 20-37) of which 13 (65%) were female. All 20 ankles had a deltoid ligament formed of four fascicles in two layers: three fascicles in the superficial layer; tibionavicular (mean dimensions 22.5 × 10.0 × 2.4 mm), tibiospring (16.6 × 6.7 × 1.9 mm) and tibiocalcaneal (23.8 × 4.6 × 1.8 mm) and a deep layer consisting of the tibiotalar fascicle, which could be divided into two parts: anterior tibiotalar (mean dimensions 10 × 5.6 × 4.1 mm) and the significantly larger posterior tibiotalar (14.2 × 13.8 × 17.5 mm, p < 0.01). There were no additional fascicles observed. The deltoid ligament complex was consistently visualised as four fascicles (tibionavicular, tibiospring, tibiocalcaneal, tibiotalar) in two layers (superficial and deep) in all 20 ankles. The posterior part of the tibiotalar fascicle was the thickest of all the fascicles in the deltoid ligament. It is, therefore, possible to accurately identify the components of the deltoid ligament, and 3T MRI can be used to assess fascicle-specific injury, which will guide treatment and rehabilitation. Level III.

Double Aortic Arch: A Comparison of Fetal CMR, Postnatal CT and Surgical Findings

In double aortic arch (DAA) one of the arches can demonstrate atretic portions postnatally, leading to diagnostic uncertainty due to overlap with isolated right aortic arch (RAA) variants. The main objective of this study is to demonstrate the morphological evolution of different DAA phenotypes from prenatal to postnatal life using 3D fetal cardiac magnetic resonance imaging (CMR) and postnatal CT/CMR imaging. 3D fetal CMR was undertaken in fetuses with suspected DAA over a six-year period (Jan 2016 - Jan 2022). All cases with surgical confirmation of DAA were retrospectively studied and morphology on fetal CMR was compared to postnatal CT/CMR and surgical findings. 32 fetuses with surgically confirmed DAA underwent fetal CMR. All demonstrated a complete DAA with left-sided arterial duct. The RAA was dominant in 30/32 (94%). Postnatal CT/CMR was undertaken at median age of 3.3months (IQR 2.0-3.9) demonstrating DAA with patency of both arches in 9/32 (28%), with 6 showing signs of coarctation of the left aortic arch (LAA). The LAA isthmus was not present on CT/CMR in 22/32(69%), the transverse arch between left carotid and left subclavian artery was not present in 1 case. Fetal CMR provides novel insights into perinatal evolution of DAA. The smaller LAA can develop coarctation or atresia related to postnatal constriction of the arterial duct, making diagnosis of DAA challenging with contrast-enhanced CT/CMR. This highlights the potentially important role for prenatal 3D vascular imaging and might improve intepretation of postnatal imaging.

Optimizing anomaly detection in 3D MRI scans: The role of ConvLSTM in medical image analysis

The analysis of Medical Images (MI), particularly the detection and classification of anomalies in 3D MRI (Magnetic Resonance Imaging) scans, plays a critical part in timely intervention and personalized therapy plans. In our paper, a comprehensive methodology for anomaly detection in 3D MRI scans of the brain is proposed that combines advanced Deep Learning (DL) techniques, including Convolutional Long Short-Term Memory (ConvLSTM) model, with efficient statistical feature extraction from segmented anomaly regions. The data collection phase utilizes three main datasets namely the Brain Tumor Segmentation Challenge (BRATS), the Federated Tumor Segmentation Challenge (FETS), and the Medical Segmentation Decathlon (MSD). The research begins with preprocessing steps including image resizing, intensity normalization, and alignment of anomaly region segmentation masks. The targeted anomaly regions within the samples are segmented using U-Net architecture and then follow statistical feature extraction procedure. Dimensionality reduction methods such as Principal Component Analysis (PCA) and Recursive Feature Elimination (RFE) are utilized to streamline the feature space. The ConvLSTM is then used to classify the anomalies using both convolution and recurrent layers to capture spatiotemporal patterns in MRI data. The model is fine-tuned and iterated for better classification performance using Adam optimizer. The statistical evaluation results with accuracy of 98.9 % showed that the designed ConvLSTM method is more suitable for clinical diagnosis and treatment design in detecting anomalies in 3D MRI images.

An Implementation of Effective CNN Model for AD Detection

This paper focuses on detecting Alzheimer’s Disease (AD). The most usual form of dementia is Alzheimer's disease, which causes permanent cause memory cell damage. Alzheimer's disease, a neurodegenerative disease, increases slowly over time. For this matter, early detection of Alzheimer's disease is important. The purpose of this work is using Magnetic Resonance Imaging (MRI) to diagnose AD. A Convolution Neural Network (CNN) model, Reset, and VGG the pre-trained learning models are used. Performing analysis and validation of layers affects the effectiveness of the model. T1-weighted MRI images are taken for preprocessing from ADNI. The Dataset images are taken from the Alzheimer's Disease Neuroimaging Initiative (ADNI). 3D MRI scans into 2D image slices shows the optimization method in the training process while achieving 96% and 94% accuracy in VGG 16 and ResNet 18 respectively. This study aims to classify AD from brain 3D MRI images and obtain better results.

Three-Dimensional MRI of Foot and Ankle: Current Perspectives and Advantages Over 2D MRI

The foot and ankle region, distinguished by its complex and intricate anatomy, is frequently susceptible to injuries, particularly in the context of sports-related activities and fall-related trauma. This article reviews the literature with respect to 3-Dimensional (3D) magnetic resonance imaging (MRI) of foot and ankle. Technical considerations and imaging and interpretation-related advantages and pitfalls will be discussed with relevant case examples. 3D MRI is advantageous over 2-Dimensional (2D) MRI in producing higher spatial resolution in a short duration of time with less partial volume artifacts and in creating structure-specific imaging using multiplanar and curved planar reconstructions. The reader will be able to apply such principles in their practice in obtaining higher-resolution volume imaging of ankle and foot with respect to ligaments, tendons, articular cartilage, plantar fascia, plantar plate, and peripheral nerves for the benefit of their patients and referring physicians.

Inter-scanner super-resolution of 3D cine MRI using a transfer-learning network for MRgRT

Objective. Deep-learning networks for super-resolution (SR) reconstruction enhance the spatial-resolution of 3D magnetic resonance imaging (MRI) for MR-guided radiotherapy (MRgRT). However, variations between MRI scanners and patients impact the quality of SR for real-time 3D low-resolution (LR) cine MRI. In this study, we present a personalized super-resolution (psSR) network that incorporates transfer-learning to overcome the challenges in inter-scanner SR of 3D cine MRI. Approach: Development of the proposed psSR network comprises two-stages: (1) a cohort-specific SR (csSR) network using clinical patient datasets, and (2) a psSR network using transfer-learning to target datasets. The csSR network was developed by training on breath-hold and respiratory-gated high-resolution (HR) 3D MRIs and their k-space down-sampled LR MRIs from 53 thoracoabdominal patients scanned at 1.5 T. The psSR network was developed through transfer-learning to retrain the csSR network using a single breath-hold HR MRI and a corresponding 3D cine MRI from 5 healthy volunteers scanned at 0.55 T. Image quality was evaluated using the peak-signal-noise-ratio (PSNR) and the structure-similarity-index-measure (SSIM). The clinical feasibility was assessed by liver contouring on the psSR MRI using an auto-segmentation network and quantified using the dice-similarity-coefficient (DSC). Results. Mean PSNR and SSIM values of psSR MRIs were increased by 57.2% (13.8–21.7) and 94.7% (0.38–0.74) compared to cine MRIs, with the reference 0.55 T breath-hold HR MRI. In the contour evaluation, DSC was increased by 15% (0.79–0.91). Average time consumed for transfer-learning was 90 s, psSR was 4.51 ms per volume, and auto-segmentation was 210 ms, respectively. Significance. The proposed psSR reconstruction substantially increased image and segmentation quality of cine MRI in an average of 215 ms across the scanners and patients with less than 2 min of prerequisite transfer-learning. This approach would be effective in overcoming cohort- and scanner-dependency of deep-learning for MRgRT.

Visibility and Size of the Pancreatic Duct and Common Bile Duct on Routine 2D MRI Sequences in Children Without Pancreaticobiliary Disease.

Background: Pancreatic duct (PD) or common bile duct (CBD) dilatation can indicate ductal pathology, but limited data describe normal pediatric duct measurements on routine 2D MR sequences. Objective: To characterize the visibility and diameter of the PD and CBD on 2D MR images in children without pancreaticobiliary disease. Methods: This retrospective study included patients who underwent abdominal MRI using a rapid protocol (comprised of noncontrast axial and coronal 2D SSFSE sequences) to assess for suspected appendicitis or ovarian torsion in the emergency department setting between January 23, 2023, and September 13, 2023, excluding patients with a pancreatic or hepatobiliary abnormality on MRI or laboratory assessment. Four radiologists independently reviewed examinations. Reviewers recorded PD visibility in each of four segments (head, neck, body, and tail) and CBD visibility, and measured PD diameter in each segment and maximal CBD diameter. Duct measurements by age were characterized by linear regression analyses. Results: The study included 177 patients [112 female, 65 female; mean age, 12.3±3.4 years (range, 5.1-17.7 years)]. The observers reported PD visibility in the head in 32.5-93.5%, neck in 18.4-71.5%, body in 22.3-69.8%, and tail in 7.3-25.7%, and in all four segments in 6.2-22.4%, of patients. Maximum PD diameter in any segment, as a mean across observers, was 1.8 mm (range across observers, 0.7-3.5 mm). Expected maximal PD diameter in any segment, in terms of the 5th and 95th percentile values of observers' mean measurements, was 1.4-2.3 mm; the prediction interval's upper limit increased from age 5 to 17 from 2.1 to 2.5 mm. All observers reported CBD visibility in all patients. Mean CBD diameter across observers was 3.1 mm (range across observers, 2.9-3.4 mm). Expected CBD diameter, in terms of the 5th and 95th percentile values of observers' mean measurements, was 2.3-4.9 mm; the prediction interval's upper limit increased from age 5 to 17 from 3.9 to 5.0 mm. Conclusion: We report expected upper limits for PD and CBD measurements on 2D MR images in children without evidence of pancreaticobiliary disease. Clinical Impact: These findings may aid radiologists' identification of pancreaticobiliary duct abnormalities on routine abdominal MRI examinations.

Deformable registration of magnetic resonance images using unsupervised deep learning in neuro-/radiation oncology

PurposeAccurate deformable registration of magnetic resonance imaging (MRI) scans containing pathologies is challenging due to changes in tissue appearance. In this paper, we developed a novel automated three-dimensional (3D) convolutional U-Net based deformable image registration (ConvUNet-DIR) method using unsupervised learning to establish correspondence between baseline pre-operative and follow-up MRI scans of patients with brain glioma.MethodsThis study involved multi-parametric brain MRI scans (T1, T1-contrast enhanced, T2, FLAIR) acquired at pre-operative and follow-up time for 160 patients diagnosed with glioma, representing the BraTS-Reg 2022 challenge dataset. ConvUNet-DIR, a deep learning-based deformable registration workflow using 3D U-Net style architecture as a core, was developed to establish correspondence between the MRI scans. The workflow consists of three components: (1) the U-Net learns features from pairs of MRI scans and estimates a mapping between them, (2) the grid generator computes the sampling grid based on the derived transformation parameters, and (3) the spatial transformation layer generates a warped image by applying the sampling operation using interpolation. A similarity measure was used as a loss function for the network with a regularization parameter limiting the deformation. The model was trained via unsupervised learning using pairs of MRI scans on a training data set (n = 102) and validated on a validation data set (n = 26) to assess its generalizability. Its performance was evaluated on a test set (n = 32) by computing the Dice score and structural similarity index (SSIM) quantitative metrics. The model’s performance also was compared with the baseline state-of-the-art VoxelMorph (VM1 and VM2) learning-based algorithms.ResultsThe ConvUNet-DIR model showed promising competency in performing accurate 3D deformable registration. It achieved a mean Dice score of 0.975 ± 0.003 and SSIM of 0.908 ± 0.011 on the test set (n = 32). Experimental results also demonstrated that ConvUNet-DIR outperformed the VoxelMorph algorithms concerning Dice (VM1: 0.969 ± 0.006 and VM2: 0.957 ± 0.008) and SSIM (VM1: 0.893 ± 0.012 and VM2: 0.857 ± 0.017) metrics. The time required to perform a registration for a pair of MRI scans is about 1 s on the CPU.ConclusionsThe developed deep learning-based model can perform an end-to-end deformable registration of a pair of 3D MRI scans for glioma patients without human intervention. The model could provide accurate, efficient, and robust deformable registration without needing pre-alignment and labeling. It outperformed the state-of-the-art VoxelMorph learning-based deformable registration algorithms and other supervised/unsupervised deep learning-based methods reported in the literature.

Development and stability analysis of carpal kinematic metrics from 4D magnetic resonance imaging.

To develop MRI-derived carpal kinematic metrics and investigating their stability. The study used a 4D MRI method to track scaphoid, lunate, and capitate movements in the wrist. A panel of 120 metrics for radial-ulnar deviation and flexion-extension was created using polynomial models of scaphoid and lunate movements relative to the capitate. Intraclass correlation coefficients (ICCs) analyzed intra- and inter-subject stability in 49 subjects, 20 with and 29 without wrist injury history. Comparable degrees of stability were observed across the two different wrist movements. Among the total 120 derived metrics, distinct subsets demonstrated high stability within each type of movement. For asymptomatic subjects, 16 out of 17 metrics with high intra-subject stability also showed high inter-subject stability. The differential analysis of ICC values for each metric between asymptomatic and symptomatic cohorts revealed specific metrics (although relatively unstable) exhibiting greater variability in the symptomatic cohort, thereby highlighting the impact of wrist conditions on the variability of kinematic metrics. The findings demonstrate the developing potential of dynamic MRI for assessing and characterizing complex carpal bone dynamics. Stability analyses of the derived kinematic metrics revealed encouraging differences between cohorts with and without wrist injury histories. Although these broad metric stability variations highlight the potential utility of this approach for analyzing carpal instability, further studies are necessary to better characterize these observations.

TACA-RNet: Tri-Axis Based Context-Aware Reverse Network for Multimodal Brain Tumor Segmentation

Brain tumor segmentation using Magnetic Resonance Imaging (MRI) is vital for clinical decision making. Traditional deep learning-based studies using convolutional neural networks have predominantly processed MRI data as two-dimensional slices, leading to the loss of contextual information. While three-dimensional (3D) convolutional layers represent an advancement, they have not fully exploited pathological information according to the three-axis nature of 3D MRI data—axial, coronal, and sagittal. Recognizing these limitations, we introduce a Tri-Axis based Context-Aware Reverse Network (TACA-RNet). This innovative approach leverages the unique 3D spatial orientations of MRI, learning crucial information on brain anatomy and pathology. We incorporated three specialized modules: a Tri-Axis Channel Reduction module for optimizing feature dimensions, a MultiScale Contextual Fusion module for aggregating multi-scale features and enhancing spatial discernment, and a 3D Axis Reverse Attention module for the precise delineation of tumor boundaries. The TACA-RNet leverages three specialized modules to enhance the understanding of tumor characteristics and spatial relationships within MRI data by fully utilizing its tri-axial structure. Validated on the Brain Tumor Segmentation Challenge 2018 and 2020 datasets, the TACA-RNet demonstrated superior performances over contemporary methodologies. This underscores the critical role of leveraging the three-axis structure of MRI to enhance segmentation accuracy.

Three-dimensional magnetic resonance imaging-based statistical shape analysis and machine learning-based prediction of patellofemoral instability.

This study performed three-dimensional (3D) magnetic resonance imaging (MRI)-based statistical shape analysis (SSA) by comparing patellofemoral instability (PFI) and normal femur models, and developed a machine learning (ML)-based prediction model. Twenty (19 patients) and 31 MRI scans (30 patients) of femurs with PFI and normal femurs, respectively, were used. Bone and cartilage segmentation of the distal femurs was performed and subsequently converted into 3D reconstructed models. The pointwise distance map showed anterior elevation of the trochlea, particularly at the central floor of the proximal trochlea, in the PFI models compared with the normal models. Principal component analysis examined shape variations in the PFI group, and several principal components exhibited shape variations in the trochlear floor and intercondylar width. Multivariate analysis showed that these shape components were significantly correlated with the PFI/non-PFI distinction after adjusting for age and sex. Our ML-based prediction model for PFI achieved a strong predictive performance with an accuracy of 0.909 ± 0.015, and an area under the curve of 0.939 ± 0.009 when using a support vector machine with a linear kernel. This study demonstrated that 3D MRI-based SSA can realistically visualize statistical results on surface models and may facilitate the understanding of complex shape features.

A modular motion compensation pipeline for prospective respiratory motion correction of multi-nuclear MR spectroscopy

Magnetic resonance (MR) acquisitions of the torso are frequently affected by respiratory motion with detrimental effects on signal quality. The motion of organs inside the body is typically decoupled from surface motion and is best captured using rapid MR imaging (MRI). We propose a pipeline for prospective motion correction of the target organ using MR image navigators providing absolute motion estimates in millimeters. Our method is designed to feature multi-nuclear interleaving for non-proton MR acquisitions and to tolerate local transmit coils with inhomogeneous field and sensitivity distributions. OpenCV object tracking was introduced for rapid estimation of in-plane displacements in 2D MR images. A full three-dimensional translation vector was derived by combining displacements from slices of multiple and arbitrary orientations. The pipeline was implemented on 3 T and 7 T MR scanners and tested in phantoms and volunteers. Fast motion handling was achieved with low-resolution 2D MR image navigators and direct implementation of OpenCV into the MR scanner’s reconstruction pipeline. Motion-phantom measurements demonstrate high tracking precision and accuracy with minor processing latency. The feasibility of the pipeline for reliable in-vivo motion extraction was shown on heart and kidney data. Organ motion was manually assessed by independent operators to quantify tracking performance. Object tracking performed convincingly on 7774 navigator images from phantom scans and different organs in volunteers. In particular the kernelized correlation filter (KCF) achieved similar accuracy (74%) as scored from inter-operator comparison (82%) while processing at a rate of over 100 frames per second. We conclude that fast 2D MR navigator images and computer vision object tracking can be used for accurate and rapid prospective motion correction. This and the modular structure of the pipeline allows for the proposed method to be used in imaging of moving organs and in challenging applications like cardiac magnetic resonance spectroscopy (MRS) or magnetic resonance imaging (MRI) guided radiotherapy.