An inter-landmark approach to 4-D shape extraction and interpretation: application to myocardial motion assessment in MRI

A deep learning-based fully automatic and clinical-ready framework for regional myocardial segmentation and myocardial ischemia evaluation

Myocardial ischemia diagnosis with CT perfusion imaging (CTP) is important in coronary artery disease management. Traditional analysis procedure is time-consuming and error-prone due to the semi-manual and operator-dependent nature. To improve the diagnostic performance, a deep learning-based, fully automatic, and clinical-ready framework was developed. Two collaborating deep learning networks including a 3D U-Net for left ventricle segmentation and a CNN for anatomical landmarks detection were trained on 276 subjects. With our processing framework, the 17-segment left ventricular model was automatically generated conformed to the clinical standard. Myocardial blood flow computed by commercial software was extracted within each segment and visualized against the bull's eye plot. The performance was validated on another 45 subjects. Coronary angiography and invasive fractional flow reserve measurements were also performed in these patients to serve as the gold standard for myocardial ischemia diagnosis. As a result, the diagnostic accuracy for our method was 81.08%, much higher than that for commercially available CTP analysis software (56.75%), and our method demonstrated a higher consistency (Kappa coefficient 0.759 vs. 0.585). Besides, the average processing time of our method was much lower (30 ± 10.5 s/subject vs. over 30 min/subject). In conclusion, the proposed deep learning-based framework could be a promising tool for assisting CTP analysis.

Regional Cardiac Motion Scoring with Multi-scale Motion-based Spatial Attention

Regional cardiac motion scoring aims to classify the motion status of each myocardium segment into one of the four categories (normal, hypokinetic, akinetic and dyskinetic) from multiple short-axis MR sequences. It is essential for prognosis and early diagnosis for various cardiac diseases. However, the complex motion procedure of myocardium and the invisible pattern differences pose great challenges, leading to low performance for automatic methods. Most existing work mitigate the task by differentiating the normal motion patterns from the abnormal ones, without fine-grained motion scoring. We propose an effective method for the task of cardiac motion scoring by connecting a bottom-up and another top-down branch with a novel motion-based spatial attention module in multi-scale space. Specifically, we use the convolution blocks for low-level feature extraction that acts as a bottom-up mechanism, and the task of optical flow for explicit motion extraction that acts as a top-down mechanism for high-level allocation of spatial attention. To this end, a newly designed Multi-scale Motion-based Spatial Attention (MMSA) mechanism is used as the pivot connecting the bottom-up part and the top-down part, and adaptively weight the low-level features according to the motion information. Experimental results on a newly constructed dataset of 1440 myocardium segments from 90 subjects demonstrate that the proposed MMSA can accurately analyze the regional myocardium motion, with accuracies of 79.3% for 4-way motion scoring, 89.0% for abnormality detection, and correlation of 0.943 for estimation of motion score index. This work has great potential for practical assessment of cardiac motion function.

A review of heart chamber segmentation for structural and functional analysis using cardiac magnetic resonance imaging

Cardiovascular magnetic resonance (CMR) has become a key imaging modality in clinical cardiology practice due to its unique capabilities for non-invasive imaging of the cardiac chambers and great vessels. A wide range of CMR sequences have been developed to assess various aspects of cardiac structure and function, and significant advances have also been made in terms of imaging quality and acquisition times. A lot of research has been dedicated to the development of global and regional quantitative CMR indices that help the distinction between health and pathology. The goal of this review paper is to discuss the structural and functional CMR indices that have been proposed thus far for clinical assessment of the cardiac chambers. We include indices definitions, the requirements for the calculations, exemplar applications in cardiovascular diseases, and the corresponding normal ranges. Furthermore, we review the most recent state-of-the art techniques for the automatic segmentation of the cardiac boundaries, which are necessary for the calculation of the CMR indices. Finally, we provide a detailed discussion of the existing literature and of the future challenges that need to be addressed to enable a more robust and comprehensive assessment of the cardiac chambers in clinical practice.

An Algorithm for the Segmentation of Highly Abnormal Hearts Using a Generic Statistical Shape Model

Statistical shape models (SSMs) have been widely employed in cardiac image segmentation. However, in conditions that induce severe shape abnormality and remodeling, such as in the case of pulmonary hypertension (PH) or hypertrophic cardiomyopathy (HCM), a single SSM is rarely capable of capturing the anatomical variability in the extremes of the distribution. This work presents a new algorithm for the segmentation of severely abnormal hearts. The algorithm is highly flexible, as it does not require a priori knowledge of the involved pathology or any specific parameter tuning to be applied to the cardiac image under analysis. The fundamental idea is to approximate the gross effect of the abnormality with a virtual remodeling transformation between the patient-specific geometry and the average shape of the reference model (e.g., average normal morphology). To define this mapping, a set of landmark points are automatically identified during boundary point search, by estimating the reliability of the candidate points. With the obtained transformation, the feature points extracted from the patient image volume are then projected onto the space of the reference SSM, where the model is used to effectively constrain and guide the segmentation process. The extracted shape in the reference space is finally propagated back to the original image of the abnormal heart to obtain the final segmentation. Detailed validation with patients diagnosed with PH and HCM shows the robustness and flexibility of the technique for the segmentation of highly abnormal hearts of different pathologies.

Automatic segmentation of the left ventricle into 17 anatomical regions in cardiac MR imaging

Analysis and characterization of anatomical segments in the left ventricle (LV) of the heart in cardiac MRI convey clinical significance. Based on the standard defined by the American Heart Association (AHA), the LV is divided into 17 anatomical segments. In this paper, we propose a novel method to automatically partition the LV into 17 segments, which allows automated analysis of these segments. Our method starts with assigning each slice with a section tag by using the papillary muscles and the LV cavity as references: basal, mid-cavity, apical and apex. It then partitions each slice into 4 or 6 segments by extracting the relevant points on the outer circle of a fitted cylinder and identifying the image orientation by using the lung as a reference. We evaluate our method on 45 patients with different cardiac conditions. The partition of mid-cavity has the best agreement with the ground truth, followed by basal and then apical sections for all groups of patients.

Effect of statistically derived fiber models on the estimation of cardiac electrical activation

Myocardial fiber orientation plays a critical role in the electrical activation and subsequent contraction of the heart. To increase the clinical potential of electrophysiological (EP) simulation for the study of cardiac phenomena and the planning of interventions, accurate personalization of the fibers is a necessary yet challenging task. Due to the difficulties associated with the in vivo imaging of cardiac fiber structure, researchers have developed alternative techniques to personalize fibers. Thus far, cardiac simulation was performed mainly based on rule-based fiber models. More recently, there has been a significant interest in data-driven and statistically derived fiber models. In particular, our predictive method in [1] allows us to estimate the unknown subject-specific fiber orientation based on the more easily available shape information. The aim of this work is to estimate the effect of using such statistical predictive models for the estimation of cardiac electrical activation times and patterns. To this end, we perform EP simulations based on a database of ten canine ex vivo diffusion tensor imaging (DTI) datasets that include normal and failing cases. To assess the strength of the fiber models under varying conditions, we consider both sinus rhythm and biventricular pacing simulations. The results show that 1) the statistically derived fibers improve the estimation of the local activation times by an average of 53.7% over traditional rule-based models, and that 2) the obtained electrical activations are consistently similar to those of the DTI-based fibers.

Regional assessment of cardiac left ventricular myocardial function via MRI statistical features

Automating the detection and localization of segmental (regional) left ventricle (LV) abnormalities in magnetic resonance imaging (MRI) has recently sparked an impressive research effort, with promising performances and a breadth of techniques. However, despite such an effort, the problem is still acknowledged to be challenging, with much room for improvements in regard to accuracy. Furthermore, most of the existing techniques are labor intensive, requiring delineations of the endo- and/or epi-cardial boundaries in all frames of a cardiac sequence. The purpose of this study is to investigate a real-time machine-learning approach which uses some image features that can be easily computed, but that nevertheless correlate well with the segmental cardiac function. Starting from a minimum user input in only one frame in a subject dataset, we build for all the regional segments and all subsequent frames a set of statistical MRI features based on a measure of similarity between distributions. We demonstrate that, over a cardiac cycle, the statistical features are related to the proportion of blood within each segment. Therefore, they can characterize segmental contraction without the need for delineating the LV boundaries in all the frames. We first seek the optimal direction along which the proposed image features are most descriptive via a linear discriminant analysis. Then, using the results as inputs to a linear support vector machine classifier, we obtain an abnormality assessment of each of the standard cardiac segments in real-time. We report a comprehensive experimental evaluation of the proposed algorithm over 928 cardiac segments obtained from 58 subjects. Compared against ground-truth evaluations by experienced radiologists, the proposed algorithm performed competitively, with an overall classification accuracy of 86.09% and a kappa measure of 0.73.

Automatic cardiac LV segmentation in MRI using modified graph cuts with smoothness and interslice constraints

PURPOSE

Magnetic resonance imaging (MRI), specifically late-enhanced MRI, is the standard clinical imaging protocol to assess cardiac viability. Segmentation of myocardial walls is a prerequisite for this assessment. Automatic and robust multisequence segmentation is required to support processing massive quantities of data.

METHODS

A generic rule-based framework to automatically segment the left ventricle myocardium is presented here. We use intensity information, and include shape and interslice smoothness constraints, providing robustness to subject- and study-specific changes. Our automatic initialization considers the geometrical and appearance properties of the left ventricle, as well as interslice information. The segmentation algorithm uses a decoupled, modified graph cut approach with control points, providing a good balance between flexibility and robustness.

RESULTS

The method was evaluated on late-enhanced MRI images from a 20-patient in-house database, and on cine-MRI images from a 15-patient open access database, both using as reference manually delineated contours. Segmentation agreement, measured using the Dice coefficient, was 0.81±0.05 and 0.92±0.04 for late-enhanced MRI and cine-MRI, respectively. The method was also compared favorably to a three-dimensional Active Shape Model approach.

CONCLUSION

The experimental validation with two magnetic resonance sequences demonstrates increased accuracy and versatility.

Automatic segmentation of cardiac MRI cines validated for long axis views

Segmentation of cardiac magnetic resonance imaging is considered an important application in clinical practice. An automatic algorithm is proposed for segmentation of both endocardial and epicardial boundaries, in long-axis views. The data consisted of 126 patients, yielding 1008 traces. Estimated clinical parameters were highly correlated to gold standard measurements. The error between the automatic tracing and the gold standard was not significantly different than the error between two manual observers. In conclusion, a tool for segmenting the myocardial boundaries in the long-axis views is proposed, which works well, as demonstrated by the validation performed using a clinical dataset.

A general framework for context-specific image segmentation using reinforcement learning

This paper presents an online reinforcement learning framework for medical image segmentation. The concept of context-specific segmentation is introduced such that the model is adaptive not only to a defined objective function but also to the user's intention and prior knowledge. Based on this concept, a general segmentation framework using reinforcement learning is proposed, which can assimilate specific user intention and behavior seamlessly in the background. The method is able to establish an implicit model for a large state-action space and generalizable to different image contents or segmentation requirements based on learning in situ. In order to demonstrate the practical value of the method, example applications of the technique to four different segmentation problems are presented. Detailed validation results have shown that the proposed framework is able to significantly reduce user interaction, while maintaining both segmentation accuracy and consistency.

Regional heart motion abnormality detection: an information theoretic approach

Tracking regional heart motion and detecting the corresponding abnormalities play an essential role in the diagnosis of cardiovascular diseases. Based on functional images, which are subject to noise and segmentation/registration inaccuracies, regional heart motion analysis is acknowledged as a difficult problem and, therefore, incorporation of prior knowledge is desirable to enhance accuracy. Given noisy data and a nonlinear dynamic model to describe myocardial motion, an unscented Kalman smoother is proposed in this study to estimate the myocardial points. Due to the similarity between the statistical information of normal and abnormal heart motions, detecting and classifying abnormality is a challenging problem. We use the Shannon's differential entropy of the distributions of potential classifier features to detect and locate regional heart motion abnormality. A naive Bayes classifier algorithm is constructed from the Shannon's differential entropy of different features to automatically detect abnormal functional regions of the myocardium. Using 174 segmented short-axis magnetic resonance cines obtained from 58 subjects (21 normal and 37 abnormal), the proposed method is quantitatively evaluated by comparison with ground truth classifications by radiologists over 928 myocardial segments. The proposed method performed significantly better than other recent methods, and yielded an accuracy of 86.5% (base), 89.4% (mid-cavity) and 84.5% (apex). The overall classification accuracy was 87.1%. Furthermore, standard kappa statistic comparisons between the proposed method and visual wall motion scoring by radiologists showed that the proposed algorithm can yield a kappa measure of 0.73.

Reinforcement learning for context aware segmentation

The ability to learn from user behavior during image segmentation to replicate the innate human ability to adapt shape delineation to contextually specific local information is an important area of study in image understanding. Current approaches to image segmentation usually incorporate specific designs, either relying on generic image features or specific prior knowledge, which usually prevent their application in different contextual settings. In this paper, a general segmentation framework based on reinforcement learning is proposed. It demonstrates how user-specific behavior can be assimilated in-situ for effective model adaptation and learning. It incorporates a two-layer reinforcement learning algorithm that constructs the model from accumulated experience during user interaction. As the algorithm learns 'pervasively' whilst the user performs manual segmentation, no additional steps are required for the training process, allowing the method to adapt and improve its accuracy as experience is acquired. Detailed validation of the method on in-vivo magnetic resonance (MR) data demonstrates the practical value of the technique in significantly reducing the level of user interaction required, whilst maintaining the overall segmentation accuracy.

Three-dimensional analysis of interventricular septal curvature from cardiac magnetic resonance images for the evaluation of patients with pulmonary hypertension

Although abnormal septal motion is a well-known sign of increased pulmonary arterial pressures, it is not routinely used to quantify the severity of pulmonary hypertension (PH). This determination relies on invasive measurements or Doppler echocardiographic estimation of right ventricular (RV) pressures, which is not always feasible or accurate in patients with PH. We hypothesized that dynamic 3D analysis of septal curvature from cardiac magnetic resonance (CMR) images may reveal differences between patients with different degrees of PH. Forty-four patients (14 controls; 30 PH patients who underwent right heart catheterization) were studied using CMR and echocardiography. CMR imaging was performed using Philips 1.5T scanner with a phased-array cardiac coil, in a retrospectively gated steady-state free precession cine mode at 30 frames per cardiac cycle. Patients were divided into 3 subgroups according to pulmonary arterial pressure. CMR images were used to reconstruct dynamic 3D left ventricular endocardial surfaces, which were analyzed to calculate septal curvature throughout the cardiac cycle. 3D curvature analysis was feasible in 88% patients. Septal curvature showed different temporal patterns in different groups. Curvature values progressively decreased with increasing severity of PH, and correlated well with invasive pressures (r-values 0.78-0.79), pulmonary vascular resistance (r = 0.83) and Doppler-derived RV peak-systolic pressure (r = 0.75). 3D analysis of septal curvature from CMR images may become a useful component in the CMR examination in patients with known or suspected PH.