Radiomics-based aortic flow profile characterization with 4D phase-contrast MRI.

4D PC MRI of the aorta has become a routinely available examination, and a multitude of single parameters have been suggested for the quantitative assessment of relevant flow features for clinical studies and diagnosis. However, clinically applicable assessment of complex flow patterns is still challenging. We present a concept for applying radiomics for the quantitative characterization of flow patterns in the aorta. To this end, we derive cross-sectional scalar parameter maps related to parameters suggested in literature such as throughflow, flow direction, vorticity, and normalized helicity. Derived radiomics features are selected with regard to their inter-scanner and inter-observer reproducibility, as well as their performance in the differentiation of sex-, age- and disease-related flow properties. The reproducible features were tested on user-selected examples with respect to their suitability for characterizing flow profile types. In future work, such signatures could be applied for quantitative flow assessment in clinical studies or disease phenotyping.

Deep learning methods for automatic evaluation of delayed enhancement-MRI. The results of the EMIDEC challenge.

A key factor for assessing the state of the heart after myocardial infarction (MI) is to measure whether the myocardium segment is viable after reperfusion or revascularization therapy. Delayed enhancement-MRI or DE-MRI, which is performed 10 min after injection of the contrast agent, provides high contrast between viable and nonviable myocardium and is therefore a method of choice to evaluate the extent of MI. To automatically assess myocardial status, the results of the EMIDEC challenge that focused on this task are presented in this paper. The challenge's main objectives were twofold. First, to evaluate if deep learning methods can distinguish between non-infarct and pathological exams, i.e. exams with or without hyperenhanced area. Second, to automatically calculate the extent of myocardial infarction. The publicly available database consists of 150 exams divided into 50 cases without any hyperenhanced area after injection of a contrast agent and 100 cases with myocardial infarction (and then with a hyperenhanced area on DE-MRI), whatever their inclusion in the cardiac emergency department. Along with MRI, clinical characteristics are also provided. The obtained results issued from several works show that the automatic classification of an exam is a reachable task (the best method providing an accuracy of 0.92), and the automatic segmentation of the myocardium is possible. However, the segmentation of the diseased area needs to be improved, mainly due to the small size of these areas and the lack of contrast with the surrounding structures.

A Collaborative Approach for the Development and Application of Machine Learning Solutions for CMR-Based Cardiac Disease Classification.

The quality and acceptance of machine learning (ML) approaches in cardiovascular data interpretation depends strongly on model design and training and the interaction with the clinical experts. We hypothesize that a software infrastructure for the training and application of ML models can support the improvement of the model training and provide relevant information for understanding the classification-relevant data features. The presented solution supports an iterative training, evaluation, and exploration of machine-learning-based multimodal data interpretation methods considering cardiac MRI data. Correction, annotation, and exploration of clinical data and interpretation of results are supported through dedicated interactive visual analytics tools. We test the presented concept with two use cases from the ACDC and EMIDEC cardiac MRI image analysis challenges. In both applications, pre-trained 2D U-Nets are used for segmentation, and classifiers are trained for diagnostic tasks using radiomics features of the segmented anatomical structures. The solution was successfully used to identify outliers in automatic segmentation and image acquisition. The targeted curation and addition of expert annotations improved the performance of the machine learning models. Clinical experts were supported in understanding specific anatomical and functional characteristics of the assigned disease classes.

Detection and analysis of cerebral aneurysms based on X-ray rotational angiography - the CADA 2020 challenge.

The Cerebral Aneurysm Detection and Analysis (CADA) challenge was organized to support the development and benchmarking of algorithms for detecting, analyzing, and risk assessment of cerebral aneurysms in X-ray rotational angiography (3DRA) images. 109 anonymized 3DRA datasets were provided for training, and 22 additional datasets were used to test the algorithmic solutions. Cerebral aneurysm detection was assessed using the F2 score based on recall and precision, and the fit of the delivered bounding box was assessed using the distance to the aneurysm. The segmentation quality was measured using the Jaccard index and a combination of different surface distance measures. Systematic errors were analyzed using volume correlation and bias. Rupture risk assessment was evaluated using the F2 score. 158 participants from 22 countries registered for the CADA challenge. The U-Net-based detection solutions presented by the community show similar accuracy compared to experts (F2 score 0.92), with a small number of missed aneurysms with diameters smaller than 3.5 mm. In addition, the delineation of these structures, based on U-Net variations, is excellent, with a Jaccard score of 0.92. The rupture risk estimation methods achieved an F2 score of 0.71. The performance of the detection and segmentation solutions is equivalent to that of human experts. The best results are obtained in rupture risk estimation by combining different image-based, morphological, and computational fluid dynamic parameters using machine learning methods. Furthermore, we evaluated the best methods pipeline, from detecting and delineating the vessel dilations to estimating the risk of rupture. The chain of these methods achieves an F2-score of 0.70, which is comparable to applying the risk prediction to the ground-truth delineation (0.71).

Comprehensive analysis of haemodynamics in patients with physiologically curved prostheses of the ascending aorta.

OBJECTIVES: This is a comprehensive analysis of haemodynamics after valve-sparing aortic root replacement (VSARR) with anatomically curved prosthesis (CP) compared to straight prosthesis (SP) and age-matched volunteers (VOL) using 4D flow MRI (time-resolved three-dimensional magnetic resonance phase-contrast imaging). METHODS: Nine patients with 90° CP, nine patients with SP, and twelve VOL were examined with 4D flow MRI. Analyses included various characteristic anatomical, qualitative and quantitative haemodynamic parameters. RESULTS: Grading of secondary flow patterns was lower in CP patients than in SP patients (P = 0.09) and more comparable to VOL, albeit not reaching statistical significance. However, it was easy to differentiate between VSARR patients and healthy volunteers: Patients more often had angular aortic arches (CP: 89%, SP: 100%; VOL: 17%; P ≤ 0.002), increased average curvature (CP: 0.17/cm [0.15, 0.18]; SP: 0.15/cm [0.14, 0.16]; VOL: 0.14/cm [0.13, 0.16]; P ≤ 0.007; values given as median [interquartile range]), and more secondary flow patterns (CP: 3 [2, 4] SP: 3 [2, 3] VOL: 2 [1, 2]; P < 0.01). Maximum circulation (CP: 142.7 cm2/s [116.1, 187.3]; SP: 101.8 cm2/s [77.7, 132.5]; VOL: 42.8cm2/s [39.3, 65.6]; P ≤ 0.002), maximum helicity density (CP: 9.6 m/s2 [9.3, 23.9]; SP: 9.7 m/s2 [8.6, 12.5]; VOL 4.9 m/s2 [4.2, 7.7]; P ≤ 0.007), and wall shear stress gradient (e.g., proximal ascending aorta CP: 0.97 N/m2 [0.54, 1.07]; SP: 1.08 N/m2 [0.74, 1.24]; VOL: 0.41 N/m2 [0.32, 0.60]; P ≤ 0.01) were increased in patients. One CP patient had a round aortic arch with physiological haemodynamic parameters. CONCLUSIONS: The restoration of physiological aortic configuration and haemodynamics was not fully achieved with the curved prostheses in our study cohort. However, there was a tendency towards improved haemodynamic conditions in the patients with curved prostheses overall but without statistical significance. A single patient with a CP and near-physiological configuration of the thoracic aorta underlines the importance of optimizing postoperative geometric conditions for allowing for physiological haemodynamics and cardiovascular energetics after VSARR.

4D flow CMR analysis comparing patients with anatomically shaped aortic sinus prostheses, tube prostheses and healthy subjects introducing the wall shear stress gradient: a case control study.

BACKGROUND: Anatomically pre-shaped sinus prostheses (SP) were developed to mimic the aortic sinus with the goal to preserve near physiological hemodynamic conditions after valve-sparing aortic root replacement. Although SP have shown more physiological flow patterns, a comparison to straight tube prosthesis and the analysis of derived quantitative parameters is lacking. Hence, this study sought to analyze differences in aortic wall shear stress (WSS) between anatomically pre-shaped SP, conventional straight tube prostheses (TP), and age-matched healthy subjects) using time-resolved 3-dimensional flow cardiovascular magnetic resonance (4D Flow CMR). Moreover, the WSS gradient was introduced and analyzed regarding its sensitivity to detect changes in hemodynamics and its dependency on the expression of secondary flow patterns. METHODS: Twelve patients with SP (12 male, 62 ± 9yr), eight patients with TP (6 male, 59 ± 9yr), and twelve healthy subjects (2 male, 55 ± 6yr) were examined at 3 T with a 4D Flow CMR sequence in this case control study. Six analysis planes were placed in the thoracic aorta at reproducible landmarks. The following WSS parameters were recorded: WSSavg (spatially averaged over the contour at peak systole), max. WSSseg (maximum segmental WSS), min. WSSseg (minimum segmental WSS) and the WSS Gradient, calculated as max. WSSseg - min. WSSseg. Kruskal-Wallis- and Mann-Whitney-U-Test were used for statistical comparison of groups. Occurrence and expression of secondary flow patterns were evaluated and correlated to WSS values using Spearman's correlation coefficient. RESULTS: In the planes bordering the prosthesis all WSS values were significantly lower in the SP compared to the TP, approaching the physiological optimum of the healthy subjects. The WSS gradient showed significantly different values in the four proximally localized contours when comparing both prostheses with healthy subjects. Strong correlations between an elevated WSS gradient and secondary flow patterns were found in the ascending aorta and the aortic arch. CONCLUSION: Overall, the SP has a positive impact on WSS, most pronounced at the site and adjacent to the prosthesis. The WSS gradient differed most obviously and the correlation of the WSS gradient with the occurrence of secondary flow patterns provides further evidence for linking disturbed flow, which was markedly increased in patients compared to healthy sub jects, to degenerative remodeling of the vascular wall.

An extensible software platform for interdisciplinary cardiovascular imaging research.

BACKGROUND AND OBJECTIVE: Cardiovascular imaging is an exponentially growing field with aspects ranging from image acquisition and analysis to disease characterization, and evaluation of therapy approaches.The transfer of innovative new technological and algorithmic solutions into clinical practice is still slow. In addition to the verification of solutions, their integration in the clinical processing workflow must be enabled for the assessment of clinical impact and risks. The goal of our software platform for cardiac image processing - CAIPI - is to support researchers from different specialties such as imaging physics, computer science, and medicine by a common extensible platform to address typical challenges and hurdles in interdisciplinary cardiovascular imaging research. It provides an integrated solution for method comparison, integrated analysis, and validation in the clinical context. The interface concept enables a combination with existing frameworks that address specific aspects of the pipeline, such as modeling (e.g., OpenCMISS, CARP) or image reconstruction (Gadgetron). METHODS: In our platform, we developed a concept for import, integration, and management of cardiac image data. The integration approach considers the spatiotemporal properties of the beating heart through a specific data model. The solution is based on MeVisLab and provides functionalities for data retrieval and storage. Two types of plugins can be added. While ToolPlugins usually provide processing algorithms such as image correction and segmentation, AnalysisPlugins enable interactive data exploration and reporting. GUI integration concepts are presented for both plugin types. We developed domain-specific reporting and visualization tools (e.g., AHA segment model) to enable validation studies by clinical experts. The platform offers plugins for calculating and reporting quantitative parameters such as cardiac function, which can be used to, e.g., evaluate the effect of processing algorithms on clinical parameters. Export functionalities include quantitative measurements to Excel, image data to PACS, and STL models to modeling and simulation tools. RESULTS: To demonstrate the applicability of this concept both for method development and clinical application, we present use cases representing different problems along the innovation chain in cardiac MR imaging. Validation of an image reconstruction method (MRI T1 mapping) Validation of an image correction method for real-time 2D-PC MRI Comparison of quantification methods for blood flow analysis Training and integration of machine learning solutions with expert annotations Clinical studies with new imaging techniques (flow measurements in the carotid arteries and peripheral veins as well as cerebral spinal fluid). CONCLUSION: The presented platform can be used in interdisciplinary teams, in which engineers or data scientists perform the method validation, followed by clinical research studies in patient collectives. The demonstrated use cases show how it enables the transfer of innovations through validation in the cardiovascular application context.

Time-resolved 3-dimensional magnetic resonance phase contrast imaging (4D Flow MRI) reveals altered blood flow patterns in the ascending aorta of patients with valve-sparing aortic root replacement.

OBJECTIVE: The aim of this study was to compare aortic flow patterns in patients after David valve-sparing aortic root replacement with physiologically shaped sinus prostheses or conventional tube grafts in healthy volunteers. METHODS: Twelve patients with sinus prostheses (55 ± 15 years), 6 patients with tube grafts (58 ± 12 years), 12 age-matched, healthy volunteers (55 ± 6 years), and 6 young, healthy volunteers (25 ± 3 years) were examined with time-resolved 3-dimensional magnetic resonance phase contrast imaging (4D Flow MRI). Primary and secondary helical, as well as vortical flow patterns, were evaluated. Aortic arch anatomy as a flow influencing factor was determined. RESULTS: Compared with volunteers, both sinus prostheses and tube grafts developed more than 4 times as many secondary flow patterns in the ascending aorta (sinus prostheses n = 1.6 ± 0.8; tube grafts n = 1.3 ± 0.6; age-matched, healthy volunteers n = 0.3 ± 0.5; young, healthy volunteers n = 0; P ≤ .012) associated with a kinking of the prosthesis itself or at its distal anastomosis. As opposed to round aortic arches in volunteers (n = 16/18), cubic or gothic-shaped arches predominated in patients (n = 16/18, P < .001). In all but 3 volunteers, 2 counter-rotating helices were confirmed in the ascending aorta and were defined as a primary flow pattern. This primary flow pattern did not develop in patients who underwent valve-sparing aortic root replacement. CONCLUSIONS: In patients after valve-sparing aortic root replacement, there was an increased number of secondary flow patterns in the ascending aorta. This seems to be related to surgically altered aortic geometry with kinking. Because flow alterations are known to affect wall shear stress, there seems to be an increased risk for vessel wall remodeling. Compared with previous 4D Flow MRI studies, primary flow patterns in the ascending aorta in healthy subjects were confirmed to be more complex. This underlines the importance of thorough examination of 4D Flow MRI data.

Time-resolved 3-dimensional magnetic resonance phase contrast imaging (4D Flow MRI) analysis of hemodynamics in valve-sparing aortic root repair with an anatomically shaped sinus prosthesis.

OBJECTIVE: The anatomically shaped sinus prosthesis (Uni-Graft W SINUS; Braun, Melsungen, Germany) used in valve-sparing aortic root replacement promises physiological hemodynamics believed to grant physiologic valve function. Using time-resolved 3-dimensional magnetic resonance phase contrast imaging (4D Flow MRI), we analyzed sinus vortex formation and transvalvular pressure gradients in patients with sinus prosthesis compared with age-matched and young healthy volunteers. METHODS: Twelve patients with sinus prosthesis (55 ± 15 years), 12 age-matched and 6 young healthy volunteers (55 ± 6 years, 25 ± 3 years, respectively) were examined at 3T with a 4D flow magnetic resonance imaging sequence. Sinus vortices visualized by streamlines and time-resolved particle paths were graded on a 4-point Likert scale. Time resolved pressure differences of the left ventricular outflow tract and the ascending aorta to a reference point in the aortic bulb as well as the transvalvular pressure gradient were evaluated. RESULTS: 4D flow visualizations revealed a propensity of the sinus prosthesis toward intermediate (50%) and large (28%) vortices compared with age-matched volunteers with small (61%) and intermediate (36%) vortices. Vortices in sinus prostheses had a similar configuration compared with those in volunteers. The peak transvalvular pressure gradient did not vary significantly between patients and age-matched volunteers (4.0 ± 0.9 mm Hg, 3.8 ± 0.7 mm Hg, P = .373), its temporal evolution resembled that of volunteers with a prolonged positive phase in patients. CONCLUSIONS: Hemodynamics closely relating to those of volunteers were confirmed in sinus prostheses, believed to grant physiological valve function. Minor differences are presumably attributed to graft compliance and temporal resolution of the acquisition. Nevertheless, long-term deterioration of valve function as it was described for straight grafts could potentially be decelerated using sinus prostheses.