Medical image analysis on left atrial LGE MRI for atrial fibrillation studies: A review

Late gadolinium enhancement magnetic resonance imaging (LGE MRI) is commonly used to visualize and quantify left atrial (LA) scars. The position and extent of LA scars provide important information on the pathophysiology and progression of atrial fibrillation (AF). Hence, LA LGE MRI computing and analysis are essential for computer-assisted diagnosis and treatment stratification of AF patients. Since manual delineations can be time-consuming and subject to intra- and inter-expert variability, automating this computing is highly desired, which nevertheless is still challenging and under-researched.This paper aims to provide a systematic review on computing methods for LA cavity, wall, scar, and ablation gap segmentation and quantification from LGE MRI, and the related literature for AF studies. Specifically, we first summarize AF-related imaging techniques, particularly LGE MRI. Then, we review the methodologies of the four computing tasks in detail and summarize the validation strategies applied in each task as well as state-of-the-art results on public datasets. Finally, the possible future developments are outlined, with a brief survey on the potential clinical applications of the aforementioned methods. The review indicates that the research into this topic is still in the early stages. Although several methods have been proposed, especially for the LA cavity segmentation, there is still a large scope for further algorithmic developments due to performance issues related to the high variability of enhancement appearance and differences in image acquisition.     

Bone and joint enhancement filtering: Application to proximal femur segmentation from uncalibrated computed tomography datasets

Methods for reliable femur segmentation enable the execution of quality retrospective studies and building of robust screening tools for bone and joint disease.An enhance-and-segment pipeline is proposed for proximal femur segmentation from computed tomography datasets. The filter is based on a scale-space model of cortical bone with properties including edge localization, invariance to density calibration, rotation invariance, and stability to noise. The filter is integrated with a graph cut segmentation technique guided through user provided sparse labels for rapid segmentation.Analysis is performed on 20 independent femurs. Rater proximal femur segmentation agreement was 0.21 mm (average surface distance), 0.98 (Dice similarity coefficient), and 2.34 mm (Hausdorff distance). Manual segmentation added considerable variability to measured failure load and volume (CV_RMS > 5%) but not density. The proposed algorithm considerably improved inter-rater reproducibility for all three outcomes (CV_RMS < 0.5%). The algorithm localized the periosteal surface accurately compared to manual segmentation but with a slight bias towards a smaller volume.Hessian-based filtering and graph cut segmentation localizes the periosteal surface of the proximal femur with comparable accuracy and improved precision compared to manual segmentation.     

AtrialJSQnet: A New framework for joint segmentation and quantification of left atrium and scars incorporating spatial and shape information

Left atrial (LA) and atrial scar segmentation from late gadolinium enhanced magnetic resonance imaging (LGE MRI) is an important task in clinical practice. The automatic segmentation is however still challenging due to the poor image quality, the various LA shapes, the thin wall, and the surrounding enhanced regions. Previous methods normally solved the two tasks independently and ignored the intrinsic spatial relationship between LA and scars. In this work, we develop a new framework, namely AtrialJSQnet, where LA segmentation, scar projection onto the LA surface, and scar quantification are performed simultaneously in an end-to-end style. We propose a mechanism of shape attention (SA) via an implicit surface projection to utilize the inherent correlation between LA cavity and scars. In specific, the SA scheme is embedded into a multi-task architecture to perform joint LA segmentation and scar quantification. Besides, a spatial encoding (SE) loss is introduced to incorporate continuous spatial information of the target in order to reduce noisy patches in the predicted segmentation. We evaluated the proposed framework on 60 post-ablation LGE MRIs from the MICCAI2018 Atrial Segmentation Challenge. Moreover, we explored the domain generalization ability of the proposed AtrialJSQnet on 40 pre-ablation LGE MRIs from this challenge and 30 post-ablation multi-center LGE MRIs from another challenge (ISBI2012 Left Atrium Fibrosis and Scar Segmentation Challenge). Extensive experiments on public datasets demonstrated the effect of the proposed AtrialJSQnet, which achieved competitive performance over the state-of-the-art. The relatedness between LA segmentation and scar quantification was explicitly explored and has shown significant performance improvements for both tasks. The code has been released via https://zmiclab.github.io/projects.html.     

3-D Ultrasound Segmentation of the Placenta Using the Random Walker Algorithm: Reliability and Agreement

Volumetric segmentation of the placenta using 3-D ultrasound is currently performed clinically to investigate correlation between organ volume and fetal outcome or pathology. Previously, interpolative or semi-automatic contour-based methodologies were used to provide volumetric results. We describe the validation of an original random walker (RW)-based algorithm against manual segmentation and an existing semi-automated method, virtual organ computer-aided analysis (VOCAL), using initialization time, inter- and intra-observer variability of volumetric measurements and quantification accuracy (with respect to manual segmentation) as metrics of success. Both semi-automatic methods require initialization. Therefore, the first experiment compared initialization times. Initialization was timed by one observer using 20 subjects. This revealed significant differences (p < 0.001) in time taken to initialize the VOCAL method compared with the RW method. In the second experiment, 10 subjects were used to analyze intra-/inter-observer variability between two observers. Bland–Altman plots were used to analyze variability combined with intra- and inter-observer variability measured by intra-class correlation coefficients, which were reported for all three methods. Intra-class correlation coefficient values for intra-observer variability were higher for the RW method than for VOCAL, and both were similar to manual segmentation. Inter-observer variability was 0.94 (0.88, 0.97), 0.91 (0.81, 0.95) and 0.80 (0.61, 0.90) for manual, RW and VOCAL, respectively. Finally, a third observer with no prior ultrasound experience was introduced and volumetric differences from manual segmentation were reported. Dice similarity coefficients for observers 1, 2 and 3 were respectively 0.84 ± 0.12, 0.94 ± 0.08 and 0.84 ± 0.11, and the mean was 0.87 ± 0.13. The RW algorithm was found to provide results concordant with those for manual segmentation and to outperform VOCAL in aspects of observer reliability. The training of an additional untrained observer was investigated, and results revealed that with the appropriate initialization protocol, results for observers with varying levels of experience were concordant. We found that with appropriate training, the RW method can be used for fast, repeatable 3-D measurement of placental volume.     

Pre-operative left atrial strain predicts post-operative atrial fibrillation in patients undergoing aortic valve replacement for aortic stenosis

Post-operative atrial fibrillation (AF) is a common and serious complication in patients undergoing aortic valve replacement (AVR). Speckle tracking echocardiography (STE) has recently enabled the quantification of longitudinal myocardial left atrial (LA) deformation dynamics. Our aim was to investigate LA preoperative mechanical function in patients undergoing AVR for aortic stenosis using STE and determine predictors of post-operative AF. 76 patients with aortic stenosis in sinus rhythm, undergoing AVR, were prospectively enrolled. Conventional echocardiographic parameters, and peak atrial longitudinal strain (PALS) were measured in all subjects the day before surgery. PALS values were obtained by averaging all segments in the 4- and 2-chamber views (global PALS). All patients received biological valve prostheses and a standard postoperative care. Postoperative AF occurred in 15 patients (19.7 %). On univariate analysis among all clinical and echocardiographic variables, global PALS showed the highest diagnostic accuracy (HR 6.55 p < 0.0001; AUC of 0.89) with a cut-off value <16.9 %, having sensitivity and specificity of 86 and 91 %, respectively, in predicting postoperative AF. LA volume indexed and E/e’ ratio had lower diagnostic accuracy (AUC 0.76 and 0.51, respectively). On multivariate analysis global PALS remains a significant predictor of postoperative AF (p < 0.0001). STE analysis of LA myocardial deformation is considered a promising tool for the evaluation of LA subclinical dysfunction in patients undergoing AVR, giving a potentially better risk stratification for the occurrence of postoperative AF.     

Mind the gap: Quantification of incomplete ablation patterns after pulmonary vein isolation using minimum path search

Pulmonary vein isolation (PVI) is a common procedure for the treatment of atrial fibrillation (AF) since the initial trigger for AF frequently originates in the pulmonary veins. A successful isolation produces a continuous lesion (scar) completely encircling the veins that stops activation waves from propagating to the atrial body. Unfortunately, the encircling lesion is often incomplete, becoming a combination of scar and gaps of healthy tissue. These gaps are potential causes of AF recurrence, which requires a redo of the isolation procedure. Late-gadolinium enhanced cardiac magnetic resonance (LGE-CMR) is a non-invasive method that may also be used to detect gaps, but it is currently a time-consuming process, prone to high inter-observer variability. In this paper, we present a method to semi-automatically identify and quantify ablation gaps. Gap quantification is performed through minimum path search in a graph where every node is a scar patch and the edges are the geodesic distances between patches. We propose the Relative Gap Measure (RGM) to estimate the percentage of gap around a vein, which is defined as the ratio of the overall gap length and the total length of the path that encircles the vein. Additionally, an advanced version of the RGM has been developed to integrate gap quantification estimates from different scar segmentation techniques into a single figure-of-merit. Population-based statistical and regional analysis of gap distribution was performed using a standardised parcellation of the left atrium. We have evaluated our method on synthetic and clinical data from 50 AF patients who underwent PVI with radiofrequency ablation. The population-based analysis concluded that the left superior PV is more prone to lesion gaps while the left inferior PV tends to have less gaps (p < .05 in both cases), in the processed data. This type of information can be very useful for the optimization and objective assessment of PVI interventions.     

Deep weakly-supervised breast tumor segmentation in ultrasound images with explicit anatomical constraints

Breast tumor segmentation is an important step in the diagnostic procedure of physicians and computer-aided diagnosis systems. We propose a two-step deep learning framework for breast tumor segmentation in breast ultrasound (BUS) images which requires only a few manual labels. The first step is breast anatomy decomposition handled by a semi-supervised semantic segmentation technique. The input BUS image is decomposed into four breast anatomical structures, namely fat, mammary gland, muscle and thorax layers. Fat and mammary gland layers are used as constrained region to reduce the search space for breast tumor segmentation. The second step is breast tumor segmentation performed in a weakly-supervised learning scenario where only image-level labels are available. Breast tumors are first recognized by a classification network and then segmented by the proposed class activation mapping and deep level set (CAM-DLS) method. For breast anatomy decomposition, the proposed framework achieves Dice similarity coefficient (DSC) of 83.0 ± 11.8%, 84.3 ± 10.0%, 80.7 ± 15.4% and 91.0 ± 11.4% for fat, mammary gland, muscle and thorax layers, respectively. For breast tumor recognition, the proposed framework achieves sensitivity of 95.8%, precision of 92.4%, specificity of 93.9%, accuracy of 94.8% and F1-score of 0.941. For breast tumor segmentation, the proposed framework achieves DSC of 77.3% and intersection-over-union (IoU) of 66.0%. In conclusion, the proposed framework could efficiently perform breast tumor recognition and segmentation simultaneously in a weakly-supervised setting with anatomical constraints.     

Evaluation of quantification methods for left arial late gadolinium enhancement based on different references in patients with atrial fibrillation

By using late gadolinium enhancement cardiac magnetic resonance (LGE-CMR) imaging, we compared left atrial late gadolinium enhancement (LA-LGE) quantification methods based on different references to characterize the left atrial wall in patients with atrial fibrillation (AF). Thirty-eight patients who underwent three-dimensional LGE-CMR imaging before catheter ablation for AF were classified into three groups depending on their clinical AF type: (1) paroxysmal AF (PAF; n = 12); (2) persistent AF (PeAF; n = 16); and (3) recurrent AF after catheter ablation (RAF; n = 10). To quantify LA-LGE on LGE-CMR imaging, we used the thresholds of 2 standard deviations (2-SD), 3-SD, 4-SD, 5-SD, or 6-SD above the mean signal from the unenhanced left ventricular myocardium, and we used the full width at half maximum (FWHM) technique, which was based on the maximum signal from the mitral valve with high signal intensity. The 6-SD threshold and FWHM techniques were statistically reproducible with an intraclass correlation coefficient >0.7. On applying the FWHM technique, the normalized LA-LGE volume by LA wall area showed a significant difference between the RAF, PeAF, and PAF groups (0.22 ± 0.04, 0.16 ± 0.06, and 0.09 ± 0.03 mL/cm², respectively) (P < 0.05). Furthermore, most of the fibrotic scarring and low-voltage tissue on the electroanatomic map corresponded well with the extent of LA-LGE. The FWHM technique based on the mitral valve can provide a reproducible quantification of LA-LGE related to AF in the thin LA wall.     

Automatic estimation of aortic and mitral valve displacements in dynamic CTA with 4D graph-cuts

The location of the mitral and aortic valves in dynamic cardiac imaging is useful for extracting functional derived parameters such as ejection fraction, valve excursions, and global longitudinal strain, and when performing anatomical structures tracking using slice following or valve intervention's planning. Completely automatic segmentation methods are still challenging tasks because of their fast movements and the different positions that prevent good visibility of the leaflets along the full cardiac cycle. In this article, we propose a processing pipeline to track the displacement of the aortic and mitral valve annuli from high-resolution cardiac four-dimensional computed tomographic angiography (4D-CTA). The proposed method is based on the dynamic separation of left ventricle, left atrium and aorta using statistical shape modeling and an energy minimization algorithm based on graph-cuts and has been evaluated on a set of 15 electrocardiography-gated 4D-CTAs. We report a mean agreement distance between manual annotations and our proposed method of 2.52±1.06 mm for the mitral annulus and 2.00±0.69 mm for the aortic valve annulus based on valve locations detected from manual anatomical landmarks. In addition, we show the effect of detecting the valvular planes on derived functional parameters (ejection fraction, global longitudinal strain, and excursions of the mitral and aortic valves).     

Liver lesion changes analysis in longitudinal CECT scans by simultaneous deep learning voxel classification with SimU-Net

The identification and quantification of liver lesions changes in longitudinal contrast enhanced CT (CECT) scans is required to evaluate disease status and to determine treatment efficacy in support of clinical decision-making. This paper describes a fully automatic end-to-end pipeline for liver lesion changes analysis in consecutive (prior and current) abdominal CECT scans of oncology patients. The three key novelties are: (1) SimU-Net, a simultaneous multi-channel 3D R2U-Net model trained on pairs of registered scans of each patient that identifies the liver lesions and their changes based on the lesion and healthy tissue appearance differences; (2) a model-based bipartite graph lesions matching method for the analysis of lesion changes at the lesion level; (3) a method for longitudinal analysis of one or more of consecutive scans of a patient based on SimU-Net that handles major liver deformations and incorporates lesion segmentations from previous analysis. To validate our methods, five experimental studies were conducted on a unique dataset of 3491 liver lesions in 735 pairs from 218 clinical abdominal CECT scans of 71 patients with metastatic disease manually delineated by an expert radiologist. The pipeline with the SimU-Net model, trained and validated on 385 pairs and tested on 249 pairs, yields a mean lesion detection recall of 0.86±0.14, a precision of 0.74±0.23 and a lesion segmentation Dice of 0.82±0.14 for lesions > 5 mm. This outperforms a reference standalone 3D R2-UNet mdel that analyzes each scan individually by ∼50% in precision with similar recall and Dice score on the same training and test datasets. For lesions matching, the precision is 0.86±0.18 and the recall is 0.90±0.15. For lesion classification, the specificity is 0.97±0.07, the precision is 0.85±0.31, and the recall is 0.86±0.23. Our new methods provide accurate and comprehensive results that may help reduce radiologists' time and effort and improve radiological oncology evaluation.     

Automatic segmentation of the lumen region in intravascular images of the coronary artery

Image assessment of the arterial system plays an important role in the diagnosis of cardiovascular diseases. The segmentation of the lumen and media-adventitia in intravascular (IVUS) images of the coronary artery is the first step towards the evaluation of the morphology of the vessel under analysis and the identification of possible atherosclerotic lesions. In this study, a fully automatic method for the segmentation of the lumen in IVUS images of the coronary artery is presented. The proposed method relies on the K-means algorithm and the mean roundness to identify the region corresponding to the potential lumen. An approach to identify and eliminate side branches on bifurcations is also proposed to delimit the area with the potential lumen regions. Additionally, an active contour model is applied to refine the contour of the lumen region. In order to evaluate the segmentation accuracy, the results of the proposed method were compared against manual delineations made by two experts in 326 IVUS images of the coronary artery. The average values of the Jaccard measure, Hausdorff distance, percentage of area difference and Dice coefficient were 0.88 ± 0.06, 0.29 ± 0.17  mm, 0.09 ± 0.07 and 0.94 ± 0.04, respectively, in 324 IVUS images successfully segmented. Additionally, a comparison with the studies found in the literature showed that the proposed method is slight better than the majority of the related methods that have been proposed. Hence, the new automatic segmentation method is shown to be effective in detecting the lumen in IVUS images without using complex solutions and user interaction.     

Characterization of the Left Atrial Vortex Flow by Two-Dimensional Transesophageal Contrast Echocardiography Using Particle Image Velocimetry

This article is the first clinical investigation of the quantitative left atrial (LA) vortex flow by two-dimensional (2-D) transesophageal contrast echocardiography (2-D-TECE) using vector particle image velocimetry (PIV). The aims of this study were to assess the feasibility of LA vortex flow analysis and to characterize and quantify the LA vortex flow in controls and in patients with atrial fibrillation (AF). Thirty-five controls and 30 patients with AF underwent transesophageal contrast echocardiography. The velocity vector was estimated by particle image velocimetry. The morphology and pulsatility of the LA vortex flow were compared between the control and AF groups. In all patients, quantitative LA vortex flow analysis was feasible. In the control group, multiple, pulsatile, compact and elliptical-shaped vortices were seen in the periphery of the LA. These vortices were persistently maintained and vectors were directed toward the atrioventricular inflow. In the AF group, a large, merged, lower pulsatile and round-shaped vortex was observed in the center of the LA. In comparisons of vortex parameters, the relative strength was significantly lower in the AF group (1.624 ± 0.501 vs. 2.105 ± 0.226, p < 0.001). It is feasible to characterize and quantify the LA vortex flow by transesophageal contrast echocardiography in patients with AF, which offers a new method to obtain additional information on LA hemodynamics. The approach has the potential for early detection of the LA dysfunction and in decisions regarding treatment strategy and guiding anticoagulation treatment in patients with AF.     

Multi-scale patch and multi-modality atlases for whole heart segmentation of MRI

A whole heart segmentation (WHS) method is presented for cardiac MRI. This segmentation method employs multi-modality atlases from MRI and CT and adopts a new label fusion algorithm which is based on the proposed multi-scale patch (MSP) strategy and a new global atlas ranking scheme. MSP, developed from the scale-space theory, uses the information of multi-scale images and provides different levels of the structural information of images for multi-level local atlas ranking. Both the local and global atlas ranking steps use the information theoretic measures to compute the similarity between the target image and the atlases from multiple modalities. The proposed segmentation scheme was evaluated on a set of data involving 20 cardiac MRI and 20 CT images. Our proposed algorithm demonstrated a promising performance, yielding a mean WHS Dice score of 0.899 ± 0.0340, Jaccard index of 0.818 ± 0.0549, and surface distance error of 1.09 ± 1.11 mm for the 20 MRI data. The average runtime for the proposed label fusion was 12.58 min.     

A new vessel segmentation algorithm for robust blood flow quantification from two‐dimensional phase‐contrast magnetic resonance images

Blood flow measurements in the ascending aorta and pulmonary artery from phase‐contrast magnetic resonance images require accurate time‐resolved vessel segmentation over the cardiac cycle. Current semi‐automatic segmentation methods often involve time‐consuming manual correction, relying on user experience for accurate results. The purpose of this study was to develop a semi‐automatic vessel segmentation algorithm with shape constraints based on manual vessel delineations for robust segmentation of the ascending aorta and pulmonary artery, to evaluate the proposed method in healthy volunteers and patients with heart failure and congenital heart disease, to validate the method in a pulsatile flow phantom experiment, and to make the method freely available for research purposes. Algorithm shape constraints were extracted from manual reference delineations of the ascending aorta (n = 20) and pulmonary artery (n = 20) and were included in a semi‐automatic segmentation method only requiring manual delineation in one image. Bias and variability (bias ± SD) for flow volume of the proposed algorithm versus manual reference delineations were 0·0 ± 1·9 ml in the ascending aorta (n = 151; seven healthy volunteers; 144 heart failure patients) and ?1·7 ± 2·9 ml in the pulmonary artery (n = 40; 25 healthy volunteers; 15 patients with atrial septal defect). Interobserver bias and variability were lower (P = 0·008) for the proposed semi‐automatic method (?0·1 ± 0·9 ml) compared to manual reference delineations (1·5 ± 5·1 ml). Phantom validation showed good agreement between the proposed method and timer‐and‐beaker flow volumes (0·4 ± 2·7 ml). In conclusion, the proposed semi‐automatic vessel segmentation algorithm can be used for efficient analysis of flow and shunt volumes in the aorta and pulmonary artery.     

Comparison of atlas-based techniques for whole-body bone segmentation

We evaluate the accuracy of whole-body bone extraction from whole-body MR images using a number of atlas-based segmentation methods. The motivation behind this work is to find the most promising approach for the purpose of MRI-guided derivation of PET attenuation maps in whole-body PET/MRI. To this end, a variety of atlas-based segmentation strategies commonly used in medical image segmentation and pseudo-CT generation were implemented and evaluated in terms of whole-body bone segmentation accuracy. Bone segmentation was performed on 23 whole-body CT/MR image pairs via leave-one-out cross validation procedure. The evaluated segmentation techniques include: (i) intensity averaging (IA), (ii) majority voting (MV), (iii) global and (iv) local (voxel-wise) weighting atlas fusion frameworks implemented utilizing normalized mutual information (NMI), normalized cross-correlation (NCC) and mean square distance (MSD) as image similarity measures for calculating the weighting factors, along with other atlas-dependent algorithms, such as (v) shape-based averaging (SBA) and (vi) Hofmann's pseudo-CT generation method. The performance evaluation of the different segmentation techniques was carried out in terms of estimating bone extraction accuracy from whole-body MRI using standard metrics, such as Dice similarity (DSC) and relative volume difference (RVD) considering bony structures obtained from intensity thresholding of the reference CT images as the ground truth. Considering the Dice criterion, global weighting atlas fusion methods provided moderate improvement of whole-body bone segmentation (DSC = 0.65 ± 0.05) compared to non-weighted IA (DSC = 0.60 ± 0.02). The local weighed atlas fusion approach using the MSD similarity measure outperformed the other strategies by achieving a DSC of 0.81 ± 0.03 while using the NCC and NMI measures resulted in a DSC of 0.78 ± 0.05 and 0.75 ± 0.04, respectively. Despite very long computation time, the extracted bone obtained from both SBA (DSC = 0.56 ± 0.05) and Hofmann's methods (DSC = 0.60 ± 0.02) exhibited no improvement compared to non-weighted IA. Finding the optimum parameters for implementation of the atlas fusion approach, such as weighting factors and image similarity patch size, have great impact on the performance of atlas-based segmentation approaches. The voxel-wise atlas fusion approach exhibited excellent performance in terms of cancelling out the non-systematic registration errors leading to accurate and reliable segmentation results. Denoising and normalization of MR images together with optimization of the involved parameters play a key role in improving bone extraction accuracy.     

Left atrial volume computed by 3D rotational angiography best predicts atrial fibrillation recurrence after circumferential pulmonary vein isolation

Left atrium (LA) size is a well-studied predictor of atrial fibrillation (AF) recurrence after pulmonary vein isolation (PVI). Yet, there is still little agreement on the best imaging technique to size the LA, and on the most appropriate sizing parameter. Volumetric assessment of LA with three-dimensional rotational angiography (3DRA LA volume index) might be a valid alternative to the commonly used transthoracic echocardiography (TTE LA volume index). The aim of our study was to investigate whether LA volume by 3DRA at the time of PVI is able to predict the risk of atrial fibrillation recurrence. We analysed 352 consecutive patients with symptomatic paroxysmal or persistent atrial fibrillation referred for PVI to our Institution. In all patients, LA volume index (LAVI) was assessed by TTE and by 3DRA. Sinus rhythm was restored after PVI in 348 patients (99%). Average TTE-LAVI and 3DRA-LAVI were 37?±?12 and 83?±?18 ml/m2, respectively. At a median follow-up of 19 (12, 24) months, 27% of patients had AF recurrence after the first PVI. At the univariate analysis, persistent AF (p?<?0.01), use of anti-arrhythmic drugs (AAD) (p?<?0.05) and 3DRA-LAVI (p?<?0.01) were significantly associated with AF recurrence. In contrast, none of the echocardiographic parameters considered, including TTE-LAVI, was associated with AF recurrence (p?=?0.29). At the multivariate analysis, 3DRA-LAVI was the only independent predictor for AF recurrence (HR 1.01 [1.00–1.03], p?=?0.017). Left atrial volume measured with 3DRA is superior to TTE assessment and to AF history in predicting atrial fibrillation recurrence after PVI.     

CAN3D: Fast 3D medical image segmentation via compact context aggregation

Direct automatic segmentation of objects in 3D medical imaging, such as magnetic resonance (MR) imaging, is challenging as it often involves accurately identifying multiple individual structures with complex geometries within a large volume under investigation. Most deep learning approaches address these challenges by enhancing their learning capability through a substantial increase in trainable parameters within their models. An increased model complexity will incur high computational costs and large memory requirements unsuitable for real-time implementation on standard clinical workstations, as clinical imaging systems typically have low-end computer hardware with limited memory and CPU resources only. This paper presents a compact convolutional neural network (CAN3D) designed specifically for clinical workstations and allows the segmentation of large 3D Magnetic Resonance (MR) images in real-time. The proposed CAN3D has a shallow memory footprint to reduce the number of model parameters and computer memory required for state-of-the-art performance and maintain data integrity by directly processing large full-size 3D image input volumes with no patches required. The proposed architecture significantly reduces computational costs, especially for inference using the CPU. We also develop a novel loss function with extra shape constraints to improve segmentation accuracy for imbalanced classes in 3D MR images. Compared to state-of-the-art approaches (U-Net3D, improved U-Net3D and V-Net), CAN3D reduced the number of parameters up to two orders of magnitude and achieved much faster inference, up to 5 times when predicting with a standard commercial CPU (instead of GPU). For the open-access OAI-ZIB knee MR dataset, in comparison with manual segmentation, CAN3D achieved Dice coefficient values of (mean = 0.87 ± 0.02 and 0.85 ± 0.04) with mean surface distance errors (mean = 0.36 ± 0.32 mm and 0.29 ± 0.10 mm) for imbalanced classes such as (femoral and tibial) cartilage volumes respectively when training volume-wise under only 12G video memory. Similarly, CAN3D demonstrated high accuracy and efficiency on a pelvis 3D MR imaging dataset for prostate cancer consisting of 211 examinations with expert manual semantic labels (bladder, body, bone, rectum, prostate) now released publicly for scientific use as part of this work.     

Assessment of left and right atrial 3D hemodynamics in patients with atrial fibrillation: a 4D flow MRI study

Atrial fibrillation (AF) is associated with embolic stroke due to thrombus formation in the left atrium (LA). Based on the relationship of atrial stasis to thromboembolism and the marked disparity in pulmonary versus systemic thromboembolism in AF, we tested the hypothesis that flow velocity distributions in the left (LA) versus right atrium (RA) in patients with would demonstrate increased stasis. Whole heart 4D flow MRI was performed in 62 AF patients (n = 33 in sinus rhythm during imaging, n = 29 with persistent AF) and 8 controls for the assessment of in vivo atrial 3D blood flow. 3D segmentation of the LA and RA geometry and normalized velocity histograms assessed atrial velocity distribution and stasis (% of atrial velocities <0.2 m/s). Atrial hemodynamics were similar for RA and LA and significantly correlated (mean velocity: r = 0.64; stasis: r = 0.55, p < 0.001). RA and LA mean and median velocities were lower in AF patients by 15–33 % and stasis was elevated by 11–19 % compared to controls. There was high inter-individual variability in LA/RA mean velocity ratio (range 0.5–1.8) and LA/RA stasis ratio (range 0.7–1.7). Patients with a history of AF and in sinus rhythm showed most pronounced differences in atrial flow (reduced mean velocities, higher stasis in the LA). While there is no systematic difference in LA versus RA flow velocity profiles, high variability was noted. Further delineation of patient specific factors and/or regional atrial effects on the LA and RA flow velocity profiles, as well as other factors such as differences in procoagulant factors, may explain the more prevalent systemic versus pulmonary thromboembolism in patients with AF.     

Myocardial scar identification based on analysis of Look–Locker and 3D late gadolinium enhanced MRI

The aim of this study is to introduce and evaluate an approach for objective and reproducible scar identification from late gadolinium enhanced (LGE) MR by analysis of LGE data with post-contrast T₁ mapping from a routinely acquired T₁ scout Look–Locker (LL) sequence. In 90 post-infarction patients, a LL sequence was acquired prior to a three-dimensional LGE sequence covering the entire left ventricle. In 60/90 patients (training set), the T₁ relaxation rates of remote myocardium and dense myocardial scar were linearly regressed to that of blood. The learned linear relationship was applied to 30/90 patients (validation set) to identify the remote myocardium and dense scar, and to normalize the LGE signal intensity to a range from 0 to 100 %. A 50 % threshold was applied to identify myocardial scar. In the validation set, two observers independently performed manual scar identification, annotated reference regions for the full-width-half-maxima (FWHM) and standard deviation (SD) method, and analyzed the LL sequence for the proposed method. Compared with the manual, FWHM, and SD methods, the proposed method demonstrated the highest inter-class correlation coefficient (0.997) and Dice overlap index (98.7 ± 1.3 %) between the two observers. The proposed method also showed excellent agreement with the gold-standard manual scar identification, with a Dice index of 89.8 ± 7.5 and 90.2 ± 6.6 % for the two observers, respectively. Combined analysis of LL and LGE sequences leads to objective and reproducible myocardial scar identification in post-infarction patients.     

Left Atrial Voltage during Atrial Fibrillation in Paroxysmal and Persistent Atrial Fibrillation Patients

Background: Left atrial (LA) endocardial voltage characteristics assessed during atrial fibrillation (AF) have not been previously compared in different AF types. This study was aimed at investigating the LA voltages and volumes in patients with paroxysmal and persistent AF. Methods: LA electroanatomic voltage maps acquired during AF were compared between consecutive patients without major structural heart disease undergoing first catheter ablation for paroxysmal AF (n = 100) or persistent AF (n = 100). The groups were comparable in baseline clinical characteristics. Results: Patients with persistent AF presented with lower median LA voltage (median 0.41, interquartile range [IQR] 0.31–0.51 mV versus median 0.99, IQR 0.47–1.56 mV; P < 0.001), and maximum LA voltage (4.07 ± 1.76 vs 6.42 ± 2.16 mV; P < 0.001). They also had a higher proportion of the LA points exhibiting voltage <0.2 mV (30 ± 20 vs 12 ± 11%; P < 0.001) and voltage 0.2–1.0 mV (55 ± 15 vs 42 ± 19%; P < 0.001). They further displayed higher LA volume/body surface area (75 ± 16 vs 58 ± 13 mL/m²; P < 0.001). In the multivariate regression model, both LA voltage (P < 10^?9) and LA volume (P < 10^?5) were significant determinants of AF type. Conclusion: Patients with persistent AF had significantly lower LA voltage compared with patients with paroxysmal AF even after adjustment for differences in indexed LA volume. LA voltage represents an independent covariate of clinical manifestation of AF. (PACE 2010; 541–548)