The success of deep learning on image classification and recognition tasks has led to new applications in diverse contexts, including the field of medical imaging. However, two properties of deep neural networks (DNNs) may limit their future use in medical applications. The first is that DNNs require a large amount of labeled training data, and the second is that the deep learning-based models lack interpretability. In this paper, we propose and investigate a data-efficient framework for the task of general medical image segmentation. We address the two aforementioned challenges by introducing domain knowledge in the form of a strong prior into a deep learning framework. This prior is expressed by a customized dynamical system. We performed experiments on two different datasets, namely JSRT and ISIC2016 (heart and lungs segmentation on chest X-ray images and skin lesion segmentation on dermoscopy images). We have achieved competitive results using the same amount of training data compared to the state-of-the-art methods. More importantly, we demonstrate that our framework is extremely data-efficient, and it can achieve reliable results using extremely limited training data. Furthermore, the proposed method is rotationally invariant and insensitive to initialization.

Background:The dorsal striatum, a nucleus in the basal ganglia, plays a key role in the execution of cognitive functions in the human brain. Recent studies have focused on how the dorsal striatum participates in a single cognitive function, whereas the specific roles of the caudate and putamen in performing multiple cognitive functions remain unclear. In this paper we conducted a meta-analysis of the relevant neuroimaging literature to understand the roles of subregions of the dorsal striatum in performing different functions.Methods:PubMed, Web of Science, and BrainMap Functional Database were searched to find original functional magnetic resonance imaging (fMRI) studies conducted on healthy adults under reward, memory, emotion, and decision-making tasks, and relevant screening criteria were formulated. Single task activation, contrast activation, and conjunction activation analyses were performed using the activation likelihood estimation (ALE) method for the coordinate-based meta-analysis to evaluate the differences and linkages.Results:In all, 112 studies were included in this meta-analysis. Analysis revealed that, of the 4 single activation tasks, reward, memory, and emotion tasks all activated the putamen more, whereas decision-making tasks activated the caudate body. Contrast analysis showed that the caudate body played an important role in the 2 cooperative activation tasks, but conjunction activation results found that more peaks appeared in the caudate head.Discussion:Different subregions of the caudate and putamen assume different roles in processing complex cognitive behaviors. Functional division of the dorsal striatum identified specific roles of 15 different subregions, reflecting differences and connections between the different subregions in performing different cognitive behaviors.

Background: Quantitative cardiovascular magnetic resonance T1-mapping is increasingly used for myocardial tissue characterization. However, the lack of standardization limits direct comparability between centers and wider roll-out for clinical use or trials. Purpose: To develop a quality assurance (QA) program assuring standardized T1 measurements for clinical use. Methods: MR phantoms manufactured in 2013 were distributed, including ShMOLLI T1-mapping and reference T1 and T2 protocols. We first studied the T1 and 12 dependency on temperature and phantom aging using phantom datasets from a single site over 4 years. Based on this, we developed a multiparametric QA model, which was then applied to 78 scans from 28 other multi-national sites. Results: T1 temperature sensitivity followed a second-order polynomial to baseline T1 values (R-2 > 0.996). Some phantoms showed aging effects, where T1 drifted up to 49% over 40 months. The correlation model based on reference T1 and T2, developed on 1004 dedicated phantom scans, predicted ShMOLLI-T1 with high consistency (coefficient of variation 1.54%), and was robust to temperature variations and phantom aging. Using the 95% confidence interval of the correlation model residuals as the tolerance range, we analyzed 390 ShMOLLI T1-maps and confirmed accurate sequence deployment in 90%(70/78) of QA scans across 28 multiple centers, and categorized the rest with specific remedial actions. Conclusions: The proposed phantom QA for T1-mapping can assure correct method implementation and protocol adherence, and is robust to temperature variation and phantom aging. This QA program circumvents the need of frequent phantom replacements, and can be readily deployed in multicenter trials. (C) 2021 The Author(s). Published by Elsevier B.V.

Abstract Background and Aims Sudden cardiac death rates are higher in patients with CKD and on haemodialysis. Hypotheses include the presence of diffuse myocardial fibrosis secondary to fluid and toxin overload. Native T1, T2 and T2* mapping through cardiac magnetic resonance (CMR) is emerging as a novel technique to quantify myocardial fibrosis. This pilot study aimed to quantify cardiac morphological change using CMR native T1, T2 and T2* mapping and correlate with autonomic provocation testing, in CKD 3b-5 and haemodialysis patients. Method Patients with stable CKD 3b and higher, and patients on haemodialysis (CKD-haemodilaysis) underwent a non-contrast CMR, which included native T1, T2, T2* mapping. Autonomic provocation testing was performed using a dipolar ECG lead, followed by 14-days of recording. Results were compared between patient groups, and T1, T2, T2* maps compared to healthy controls using the student t test and Kruskal-Wallis tests. Results Nine CKD, eight haemodialysis and seven control patients were recruited (Table 1). Of the late-stage CKD patients, three were stage 3b, four were stage 4 and two were stage 5. There were no significant differences between the two patient groups in baseline characteristics (Table 1). There were no significant differences between CKD and CKD-haemodialysis patients in left ventricular end-diastolic volume index, left ventricular end-systolic volume index, right ventricular end-diastolic volume index, right ventricular end-systolic volume index, ejection fraction, and left ventricular mass index (71.1±15.2 vs. 80.51 ±21.9 ml/m2, p=0.316; 24.4±7.09 vs. 34.4±19.4 ml/m2, p=0.171; 67.11 ± 14.9 vs. 75.5±23.4 ml/m2, p=0.386; 22.2±4.87 vs. 23.9±9.93 ml/m2, p=0.663; 65.8±6.34 vs. 59.5±12.4 %, p=0.200; 48.4±8.60 vs. 50.5±11.0 g/m2, p=0.673). T1 and T2 were significantly increased in CKD and CKD-haemodialysis patients compared to healthy controls (1259±57.7 vs. 1204±22.3 ms, p=0.038 and 49.1±4.74 vs. 42.0±2.79 ms, p=0.034). There was no difference in T2* star (32.8±7.59 vs. 28.8±3.77, p=0.291). There was no significant difference in native T1, T2 and T2* times between CKD and CKD-haemodialysis patients (1247±66.7 vs. 1273±45.7, p=0.361; 49.1±5.22 vs. 49.0±4.49, p=0.960; 34.1±7.57 vs. 31.3±7.81, p=0.769). Mean percentage change of HR in CKD patients from lying to sitting to standing was 4.51%±6.66 and 11.5%±11.8 respectively. Mean percentage change of HR in CKD-haemodialysis from lying to sitting to standing was 2.15%±6.30 and 6.0%±4.45 respectively. There were no significant differences in postural HR variability between CKD and CKD-haemodialysis patients (p=0.478 and p=0.237). Conclusion In late stage CKD, cardiac volumes, mass, ejection fraction and native T1, T2 and T2* are comparable to those of patients on long-term haemodialysis. However native T1 and T2 times are significantly elevated in later stage CKD and haemodialysis, compared to healthy controls. Heart rate changes over postural provocation are comparable between CKD and CKD-haemodialysis patients, although autonomic response is reduced compared to previously published data in healthy controls. Processes that drive myocardial fibrosis may start earlier in CKD pathogenesis.

Background: Hypertrophic cardiomyopathy (HCM) has a variable phenotypic expression.Numerous studies have noted that individuals of African/Afro-Caribbean origin (black) and Caucasians (white) demonstrate marked differences on ECG and echocardiography.However, there is a paucity of literature on cardiac MRI (CMR) findings of ethnic differences in HCM.Purpose: Assess if phenotypic differences on CMR exist between black and white individuals with HCM.Methods: Patients with HCM were recruited from a dedicated inherited cardiac conditions clinic if they were NYHA class 1, had a blood pressure ≤140/90mmHg, no evidence of LV outflow tract obstruction and were not known to have ischemic heart disease (n = 119, males = 99, white = 76, aged 16-65 years).Sequences were performed on a 3T Phillips scanner, using standard techniques and analysed using semi-automated software (MRI extended workspace 2.6.3.3,Phillips and Circle, Cardiovascular Imaging Inc, release 5.6.1,Calgary, Canada).Native T1 was calculated using ShMOLLI sequences for the basal, mid and apical septal segments and the mid-ventricular short axis slice.Results: Comparison of left ventricular (LV) mass between ethnicities demonstrated higher LV mass in black compared to white HCM patients (p < 0.001, absolute values and normalised for body surface area).Mean LV mass (normalised) for black and white HCM patients was 91.21g/m² and 74.29g/m² respectively.Strain analysis demonstrated a 7% lower peak global radial strain percentage and a 3% lower peak global circumferential strain percentage in black HCM patients compared to white HCM patients (p = 0.010, p = 0.001 respectively).Black HCM patients demonstrated less late gadolinium enhancement (LGE) than white HCM patients by mass (grams) and by percentage of myocardium (auto-threshold using 6 standard deviations) (p = 0.005, p < 0.001 respectively).Analysis of native T1 demonstrated that black HCM patients had a lower native T1 in the basal septum compared to white HCM patients (p = 0.006).There were no statistically significant ethnic differences in maximal LV wall thickness, LV end diastolic volume, left atrial area, LV stroke volume, LV ejection fraction, or peak global longitudinal strain percentage. Conclusion:This is one of the first CMR studies assessing for ethnic differences in tissue characterisation in HCM between black and white patients.Findings from this study found black HCM patients had greater LV mass, reduced peak global radial and circumferential strain percentage, lower quantity of LGE, and lower native T1 values compared to white HCM patients.

Background The combination of multi-channel data is a critical step for the imaging of phase and susceptibility contrast in magnetic resonance imaging (MRI). Magnitude-weighted phase combination methods often produce noise and aliasing artifacts in the magnitude images at accelerated imaging sceneries. To address this issue, an optimal coil combination method through deconvolution in k-space is proposed in this paper. Methods The proposed method firstly employs the sum-of-squares and phase aligning method to yield a complex reference coil image which is then used to calculate the coil sensitivity and its Fourier transform. Then, the coil k-space combining weights is computed, taking into account the truncated frequency data of coil sensitivity and the acquired k-space data. Finally, combining the coil k-space data with the acquired weights generates the k-space data of proton distribution, with which both phase and magnitude information can be obtained straightforwardly. Both phantom and in vivo imaging experiments were conducted to evaluate the performance of the proposed method. Results Compared with magnitude-weighted method and MCPC-C, the proposed method can alleviate the phase cancellation in coil combination, resulting in a less wrapped phase. Conclusions The proposed method provides an effective and efficient approach to combine multiple coil image in parallel MRI reconstruction, and has potential to benefit routine clinical practice in the future.

Precise segmentation of the left ventricle (LV) within cardiac MRI images is a prerequisite for the quantitative measurement of heart function. However, this task is challenging due to the limited availability of labeled data and motion artifacts from cardiac imaging. In this work, we present an iterative segmentation algorithm for LV delineation. By coupling deep learning with a novel dynamic-based labeling scheme, we present a new methodology where a policy model is learned to guide an agent to travel over the image, tracing out a boundary of the ROI – using the magnitude difference of the Poincaré map as a stopping criterion. Our method is evaluated on two datasets, namely the Sunnybrook Cardiac Dataset (SCD) and data from the STACOM 2011 LV segmentation challenge. Our method outperforms the previous research over many metrics. In order to demonstrate the transferability of our method we present encouraging results over the STACOM 2011 data, when using a model trained on the SCD dataset.

PurposeTo improve liver mapping by incorporating adaptive neighborhood regularization into pixel‐wise curve fitting.MethodsMagnetic resonance imaging mapping remains challenging because of the serial images with low signal‐to‐noise ratio. In this study, we proposed to exploit the neighboring pixels as regularization terms and adaptively determine the regularization parameters according to the interpixel signal similarity. The proposed algorithm, called the pixel‐wise curve fitting with adaptive neighborhood regularization (PCANR), was compared with the conventional nonlinear least squares (NLS) and nonlocal means filter‐based NLS algorithms on simulated, phantom, and in vivo data.ResultsVisually, the PCANR algorithm generates maps with significantly reduced noise and well‐preserved tiny structures. Quantitatively, the PCANR algorithm produces maps with lower root mean square errors at varying values and signal‐to‐noise‐ratio levels compared with the NLS and nonlocal means filter‐based NLS algorithms. For the high values under low signal‐to‐noise‐ratio levels, the PCANR algorithm outperforms the NLS and nonlocal means filter‐based NLS algorithms in the accuracy and precision, in terms of mean and standard deviation of measurements in selected region of interests, respectively.ConclusionsThe PCANR algorithm can reduce the effect of noise on liver mapping, and the improved measurement precision will benefit the assessment of hepatic iron in clinical practice. Magn Reson Med 80:792–801, 2018. © 2018 International Society for Magnetic Resonance in Medicine.

Revealing the relationship between dysfunctional genes in blood and brain tissues from patients with Alzheimer's Disease (AD) will help us to understand the pathology of this disease. In this study, we conducted the first such large systematic analysis to identify differentially expressed genes (DEGs) in blood samples from 245 AD cases, 143 mild cognitive impairment (MCI) cases, and 182 healthy control subjects, and then compare these with DEGs in brain samples. We evaluated our findings using two independent AD blood datasets and performed a gene-based genome-wide association study to identify potential novel risk genes. We identified 789 and 998 DEGs common to both blood and brain of AD and MCI subjects respectively, over 77% of which had the same regulation directions across tissues and disease status, including the known ABCA7, and the novel TYK2 and TCIRG1. A machine learning classification model containing NDUFA1, MRPL51, and RPL36AL, implicating mitochondrial and ribosomal function, was discovered which discriminated between AD patients and controls with 85.9% of area under the curve and 78.1% accuracy (sensitivity = 77.6%, specificity = 78.9%). Moreover, our findings strongly suggest that mitochondrial dysfunction, NF-κB signalling and iNOS signalling are important dysregulated pathways in AD pathogenesis.

Liver R2* mapping is often degraded by the low signal-to-noise ratio (SNR) especially in the presence of severe iron. This study aims to improve liver R2* mapping at low SNRs by averaging decay curves before the process of curve-fitting. Independently filtering echo images by nonlocal means (NLM) demonstrated improved quality of R2* mapping, but may introduce new errors due to the nonlinear nature of the NLM filter, during which the averaging weights may vary with different image contents at multiple echo times. In addition, the image denoising effect of the NLM may decline when no sufficient similar patches are available. To overcome these drawbacks, we proposed to filter decay curves instead of images. In this novel scheme, decay curves were averaged in a local window, each with a weight assigned according to the curve-similarity measured by the distance between one of the neighboring curves and the targeted one. The proposed method was tested on simulated, phantom and patient data. The results demonstrate that the proposed method can provide more accurate R2* mapping compared with the NLM algorithm, and hence has the potential to improve diagnosis and therapy in patients with liver iron.

Heart failure related to cardiac siderosis remains a major cause of death in transfusion dependent anaemias. Replacement fibrosis has been reported as causative of heart failure in siderotic cardiomyopathy in historical reports, but these findings do not accord with the reversible nature of siderotic heart failure achievable with intensive iron chelation.

Cardiovascular Magnetic Resonance (CMR), a non-invasive and nonionizing imaging technique, plays a major role in research and clinical cardiology. The strength of CMR lies in its high temporal resolution, superior contrast, and unique tissue characterization capabilities. Contrast agents have been used to improve sensitivity and specificity of CMR in detecting and evaluating various pathologies. Much effort has been made to develop more efficient contrast reagents to detect cardiovascular diseases at an asymptomatic stage, which has led to a plethora of products in animal studies. However, very few of the developed contrast agents are currently approved for human use. Major obstacles are high dosages, toxicity, body clearance rate and long-term immunogenicity. In this review, we critically assess recent developments in the field of the contrast agents for CMR, highlighting both benefits and current drawbacks. A clearer insight regarding the challenges facing the development of improved contrast agents may help collaborative work to enhance images contrast, decrease toxicity and accelerate their translation into clinical use.

Precise segmentation of the left ventricle (LV) within cardiac MRI images is a prerequisite for the quantitative measurement of heart function. However, this task is challenging due to the limited availability of labeled data and motion artifacts from cardiac imaging. In this work, we present an iterative segmentation algorithm for LV delineation. By coupling deep learning with a novel dynamic-based labeling scheme, we present a new methodology where a policy model is learned to guide an agent to travel over the the image, tracing out a boundary of the ROI -- using the magnitude difference of the Poincare map as a stopping criterion. Our method is evaluated on two datasets, namely the Sunnybrook Cardiac Dataset (SCD) and data from the STACOM 2011 LV segmentation challenge. Our method outperforms the previous research over many metrics. In order to demonstrate the transferability of our method we present encouraging results over the STACOM 2011 data, when using a model trained on the SCD dataset.

Parametric mapping techniques provide a non-invasive tool for quantifying tissue alterations in myocardial disease in those eligible for cardiovascular magnetic resonance (CMR). Parametric mapping with CMR now permits the routine spatial visualization and quantification of changes in myocardial composition based on changes in T1, T2, and T2*(star) relaxation times and extracellular volume (ECV). These changes include specific disease pathways related to mainly intracellular disturbances of the cardiomyocyte (e.g., iron overload, or glycosphingolipid accumulation in Anderson-Fabry disease); extracellular disturbances in the myocardial interstitium (e.g., myocardial fibrosis or cardiac amyloidosis from accumulation of collagen or amyloid proteins, respectively); or both (myocardial edema with increased intracellular and/or extracellular water). Parametric mapping promises improvements in patient care through advances in quantitative diagnostics, inter- and intra-patient comparability, and relatedly improvements in treatment. There is a multitude of technical approaches and potential applications. This document provides a summary of the existing evidence for the clinical value of parametric mapping in the heart as of mid 2017, and gives recommendations for practical use in different clinical scenarios for scientists, clinicians, and CMR manufacturers.

There is a need for improved worldwide access to tissue iron quantification using T2* cardiovascular magnetic resonance (CMR). One route to facilitate this would be simple in-line T2* analysis widely available on MR scanners. We therefore compared our clinically validated and established T2* method at Royal Brompton Hospital (RBH T2*) against a novel work-in-progress (WIP) sequence with in-line T2* measurement from Siemens (WIP T2*).

T2* magnetic resonance of tissue iron concentration has improved the outcome of transfusion dependant anaemia patients. Clinical evaluation is performed at 1.5 T but scanners operating at 3 T are increasing in numbers. There is a paucity of data on the relative merits of iron quantification at 3 T vs 1.5 T.

Introduction Current intervention criteria for the thoracic aorta concentrate on size. However, the complexity of aortic disease is not fully exposed by aortic dimensions alone, and morbidity or mortality can occur before intervention thresholds are reached. Computational fluid dynamics (CFD) is a non-invasive approach to quantify haemodynamics in assessment of aneurysms and rupture risk. Wall shear stress (WSS) measuring viscous shearing forces on the endothelium, and oscillatory shear index (OSI) measuring disturbed flow, are a pathophysiological stimulus to gene expression, extracellular-matrix remodelling, and aortic wall thinning. We aimed to evaluate the efficacy of a new patient-specific approach to CFD of the thoracic aorta, and its functional and haemodynamic indices in assessment of aortic pathology. Methods 45 subjects were divided into 5 groups: Volunteers, AR-TAV, AS-TAV, AS-BAV(RL), AS-BAV(RN), where AR=aortic regurgitation, AS=aortic stenosis, TAV=tricuspid aortic valve, BAV=bicuspid aortic valve, RL=right-left cusp fusion, RN=right-non cusp fusion. Subjects underwent magnetic resonance angiography, with phase-contrast MRI at the sino-tubular junction to define patient-specific inflow velocity profiles. Three-dimensional aorta models were constructed from MRA data and discretized to form a finite element mesh. The 3D velocity profile from PC-MRI was mapped onto the inflow mesh, allowing prescription of patient-specific inflow boundary conditions. Blood pressure, cardiac output, and cross-sectional area of each vessel were processed to assign outflow boundary conditions to arch vessels and descending aorta. Results CFD enabled measurement of WSS throughout the thoracic aorta. WSS was significantly elevated in aortic stenosis, highest in AS-BAV(RN) (mean WSS=37.1 ± 4.0 dyn/cm2, compared to 19.9 ± 1.9 dyn/cm2 for AS-BAV(RL), 25.7 ± 1.2 dyn/cm2 in AS-TAV, 12.3 ± 3.4 dyn/cm2 in AR-TAV, and 9.9 ± 5.4 dyn/cm2 in healthy volunteers, p < 0.05). Aortic stenosis patients displayed asymmetrical WSS distributions, the greater curvature experiencing the highest WSS. OSI was lower in bicuspid right-non fusion (p < 0.05). Eccentricity of flow was higher in bicuspid patients (Flowasymmetry= 84.1 ± 5.4%, compared to 28.1 ± 21.5 for tricuspids, p < 0.05). Helicity of flow was assessed by the Helical Flow Index (HFI), which was higher in bicuspid right-left fusion (HFIsystole= 0.39 ± 0.04, compared to 0.28 ± 0.03 for all others, p < 0.05). Conclusions BAV displays eccentric flow with high helicity. Presence of AS, particularly in BAV-RN led to higher WSS and lower OSI in the greater curvature of the ascending aorta. Patient-specific CFD provides non-invasive functional assessment of the thoracic aorta, and enables development of a personalized approach to diagnosis and management of aortic disease beyond traditional guidelines.

OBJECTIVES:The complexity of aortic disease is not fully exposed by aortic dimensions alone, and morbidity or mortality can occur before intervention thresholds are met. Patient-specific computational fluid dynamics (CFD) were used to assess the effect of different aortic valve morphologies on velocity profiles, flow patterns, helicity, wall shear stress (WSS), and oscillatory shear index (OSI) in the thoracic aorta. METHODS:A total of 45 subjects were divided into 5 groups: volunteers, aortic regurgitation-tricuspid aortic valve (AR-TAV), aortic stenosis-tricuspid aortic valve (AS-TAV), aortic stenosis-bicuspid aortic valve right-left cusp fusion (BAV[RL]), and aortic stenosis-right-non cusp fusion (AS-BAV[RN]). Subjects underwent magnetic resonance angiography, with phase-contrast magnetic resonance imaging at the sino-tubular junction to define patient-specific inflow velocity profiles. Hemodynamic recordings were used alongside magnetic resonance imaging angiographic data to run patient-specific CFD. RESULTS:The BAV groups had larger mid-ascending aorta diameters (P < .05). Ascending aorta flow was more eccentric in BAV (flow asymmetry = 78.9% ± 6.5% for AS-BAV(RN), compared with 4.7% ± 2.1% for volunteers, P < .05). Helicity was greater in AS-BAV(RL) (P < .05). Mean WSS was elevated in AS groups, greatest in AS-BAV(RN) (37.1 ± 4.0 dyn/cm2, compared with 9.8 ± 5.4 for volunteers, P < .05). The greater curvature of the ascending aorta experienced highest WSS and lowest OSI in AS patients, most significant in AS-BAV(RN) (P < .05). CONCLUSIONS:BAV displays eccentric flow with high helicity. The presence of AS, particularly in BAV-RN, led to greater WSS and lower OSI in the greater curvature of the ascending aorta. Patient-specific CFD provides noninvasive functional assessment of the thoracic aorta, and may enable development of a personalized approach to diagnosis and management of aortic disease beyond traditional guidelines.

Myocardial black blood (BB) T2* relaxometry at 1.5T provides robust, reproducible and calibrated non-invasive assessment of cardiac iron burden. In vitro data has shown that like T2*, novel native Modified Look-Locker Inversion recovery (MOLLI) T1 shortens with increasing tissue iron. The relative merits of T1 and T2* are largely unexplored. We compared the established 1.5T BB T2* technique against native T1 values at 1.5T and 3T in iron overload patients and in normal volunteers.