Bootstrap Representation Learning for Segmentation on Medical Volumes and Sequences

In this work, we propose a novel straightforward method for medical volume and sequence segmentation with limited annotations. To avert laborious annotating, the recent success of self-supervised learning(SSL) motivates the pre-training on unlabeled data. Despite its success, it is still challenging to adapt typical SSL methods to volume/sequence segmentation, due to their lack of mining on local semantic discrimination and rare exploitation on volume and sequence structures. Based on the continuity between slices/frames and the common spatial layout of organs across volumes/sequences, we introduced a novel bootstrap self-supervised representation learning method by leveraging the predictable possibility of neighboring slices. At the core of our method is a simple and straightforward dense self-supervision on the predictions of local representations and a strategy of predicting locals based on global context, which enables stable and reliable supervision for both global and local representation mining among volumes. Specifically, we first proposed an asymmetric network with an attention-guided predictor to enforce distance-specific prediction and supervision on slices within and across volumes/sequences. Secondly, we introduced a novel prototype-based foreground-background calibration module to enhance representation consistency. The two parts are trained jointly on labeled and unlabeled data. When evaluated on three benchmark datasets of medical volumes and sequences, our model outperforms existing methods with a large margin of 4.5\% DSC on ACDC, 1.7\% on Prostate, and 2.3\% on CAMUS. Intensive evaluations reveals the effectiveness and superiority of our method.

Joint Landmark and Structure Learning for Automatic Evaluation of Developmental Dysplasia of the Hip

The ultrasound (US) screening of the infant hip is vital for the early diagnosis of developmental dysplasia of the hip (DDH). The US diagnosis of DDH refers to measuring alpha and beta angles that quantify hip joint development. These two angles are calculated from key anatomical landmarks and structures of the hip. However, this measurement process is not trivial for sonographers and usually requires a thorough understanding of complex anatomical structures. In this study, we propose a multi-task framework to learn the relationships among landmarks and structures jointly and automatically evaluate DDH. Our multi-task networks are equipped with three novel modules. Firstly, we adopt Mask R-CNN as the basic framework to detect and segment key anatomical structures and add one landmark detection branch to form a new multi-task framework. Secondly, we propose a novel shape similarity loss to refine the incomplete anatomical structure prediction robustly and accurately. Thirdly, we further incorporate the landmark-structure consistent prior to ensure the consistency of the bony rim estimated from the segmented structure and the detected landmark. In our experiments, 1,231 US images of the infant hip from 632 patients are collected, of which 247 images from 126 patients are tested. The average errors in alpha and beta angles are 2.221 degrees and 2.899 degrees. About 93% and 85% estimates of alpha and beta angles have errors less than 5 degrees, respectively. Experimental results demonstrate that the proposed method can accurately and robustly realize the automatic evaluation of DDH, showing great potential for clinical application.

Agent with Warm Start and Adaptive Dynamic Termination for Plane Localization in 3D Ultrasound

Accurate standard plane (SP) localization is the fundamental step for prenatal ultrasound (US) diagnosis. Typically, dozens of US SPs are collected to determine the clinical diagnosis. 2D US has to perform scanning for each SP, which is time-consuming and operator-dependent. While 3D US containing multiple SPs in one shot has the inherent advantages of less user-dependency and more efficiency. Automatically locating SP in 3D US is very challenging due to the huge search space and large fetal posture variations. Our previous study proposed a deep reinforcement learning (RL) framework with an alignment module and active termination to localize SPs in 3D US automatically. However, termination of agent search in RL is important and affects the practical deployment. In this study, we enhance our previous RL framework with a newly designed adaptive dynamic termination to enable an early stop for the agent searching, saving at most 67% inference time, thus boosting the accuracy and efficiency of the RL framework at the same time. Besides, we validate the effectiveness and generalizability of our algorithm extensively on our in-house multi-organ datasets containing 433 fetal brain volumes, 519 fetal abdomen volumes, and 683 uterus volumes. Our approach achieves localization error of 2.52mm/10.26 degrees, 2.48mm/10.39 degrees, 2.02mm/10.48 degrees, 2.00mm/14.57 degrees, 2.61mm/9.71 degrees, 3.09mm/9.58 degrees, 1.49mm/7.54 degrees for the transcerebellar, transventricular, transthalamic planes in fetal brain, abdominal plane in fetal abdomen, and mid-sagittal, transverse and coronal planes in uterus, respectively. Experimental results show that our method is general and has the potential to improve the efficiency and standardization of US scanning.

Style-invariant Cardiac Image Segmentation with Test-time Augmentation

Deep models often suffer from severe performance drop due to the appearance shift in the real clinical setting. Most of the existing learning-based methods rely on images from multiple sites/vendors or even corresponding labels. However, collecting enough unknown data to robustly model segmentation cannot always hold since the complex appearance shift caused by imaging factors in daily application. In this paper, we propose a novel style-invariant method for cardiac image segmentation. Based on the zero-shot style transfer to remove appearance shift and test-time augmentation to explore diverse underlying anatomy, our proposed method is effective in combating the appearance shift. Our contribution is three-fold. First, inspired by the spirit of universal style transfer, we develop a zero-shot stylization for content images to generate stylized images that appearance similarity to the style images. Second, we build up a robust cardiac segmentation model based on the U-Net structure. Our framework mainly consists of two networks during testing: the ST network for removing appearance shift and the segmentation network. Third, we investigate test-time augmentation to explore transformed versions of the stylized image for prediction and the results are merged. Notably, our proposed framework is fully test-time adaptation. Experiment results demonstrate that our methods are promising and generic for generalizing deep segmentation models.

FetusMap: Fetal Pose Estimation in 3D Ultrasound

The 3D ultrasound (US) entrance inspires a multitude of automated prenatal examinations. However, studies about the structuralized description of the whole fetus in 3D US are still rare. In this paper, we propose to estimate the 3D pose of fetus in US volumes to facilitate its quantitative analyses in global and local scales. Given the great challenges in 3D US, including the high volume dimension, poor image quality, symmetric ambiguity in anatomical structures and large variations of fetal pose, our contribution is three-fold. (i) This is the first work about 3D pose estimation of fetus in the literature. We aim to extract the skeleton of whole fetus and assign different segments/joints with correct torso/limb labels. (ii) We propose a self-supervised learning (SSL) framework to finetune the deep network to form visually plausible pose predictions. Specifically, we leverage the landmark-based registration to effectively encode case-adaptive anatomical priors and generate evolving label proxy for supervision. (iii) To enable our 3D network perceive better contextual cues with higher resolution input under limited computing resource, we further adopt the gradient check-pointing (GCP) strategy to save GPU memory and improve the prediction. Extensively validated on a large 3D US dataset, our method tackles varying fetal poses and achieves promising results. 3D pose estimation of fetus has potentials in serving as a map to provide navigation for many advanced studies.

Agent with Warm Start and Active Termination for Plane Localization in 3D Ultrasound

Standard plane localization is crucial for ultrasound (US) diagnosis. In prenatal US, dozens of standard planes are manually acquired with a 2D probe. It is time-consuming and operator-dependent. In comparison, 3D US containing multiple standard planes in one shot has the inherent advantages of less user-dependency and more efficiency. However, manual plane localization in US volume is challenging due to the huge search space and large fetal posture variation. In this study, we propose a novel reinforcement learning (RL) framework to automatically localize fetal brain standard planes in 3D US. Our contribution is two-fold. First, we equip the RL framework with a landmark-aware alignment module to provide warm start and strong spatial bounds for the agent actions, thus ensuring its effectiveness. Second, instead of passively and empirically terminating the agent inference, we propose a recurrent neural network based strategy for active termination of the agent's interaction procedure. This improves both the accuracy and efficiency of the localization system. Extensively validated on our in-house large dataset, our approach achieves the accuracy of 3.4mm/9.6{\deg} and 2.7mm/9.1{\deg} for the transcerebellar and transthalamic plane localization, respectively. Ourproposed RL framework is general and has the potential to improve the efficiency and standardization of US scanning.

Segmentation of Multimodal Myocardial Images Using Shape-Transfer GAN

Myocardium segmentation of late gadolinium enhancement (LGE) Cardiac MR images is important for evaluation of infarction regions in clinical practice. The pathological myocardium in LGE images presents distinctive brightness and textures compared with the healthy tissues, making it much more challenging to be segment. Instead, the balanced-Steady State Free Precession (bSSFP) cine images show clearly boundaries and can be easily segmented. Given this fact, we propose a novel shape-transfer GAN for LGE images, which can 1) learn to generate realistic LGE images from bSSFP with the anatomical shape preserved, and 2) learn to segment the myocardium of LGE images from these generated images. It's worth to note that no segmentation label of the LGE images is used during this procedure. We test our model on dataset from the Multi-sequence Cardiac MR Segmentation Challenge. The results show that the proposed Shape-Transfer GAN can achieve accurate myocardium masks of LGE images.

Cardiac Motion Scoring with Segment- and Subject-level Non-Local Modeling

Motion scoring of cardiac myocardium is of paramount importance for early detection and diagnosis of various cardiac disease. It aims at identifying regional wall motions into one of the four types including normal, hypokinetic, akinetic, and dyskinetic, and is extremely challenging due to the complex myocardium deformation and subtle inter-class difference of motion patterns. All existing work on automated motion analysis are focused on binary abnormality detection to avoid the much more demanding motion scoring, which is urgently required in real clinical practice yet has never been investigated before. In this work, we propose Cardiac-MOS, the first powerful method for cardiac motion scoring from cardiac MR sequences based on deep convolution neural network. Due to the locality of convolution, the relationship between distant non-local responses of the feature map cannot be explored, which is closely related to motion difference between segments. In Cardiac-MOS, such non-local relationship is modeled with non-local neural network within each segment and across all segments of one subject, i.e., segment- and subject-level non-local modeling, and lead to obvious performance improvement. Besides, Cardiac-MOS can effectively extract motion information from MR sequences of various lengths by interpolating the convolution kernel along the temporal dimension, therefore can be applied to MR sequences of multiple sources. Experiments on 1440 myocardium segments of 90 subjects from short axis MR sequences of multiple lengths prove that Cardiac-MOS achieves reliable performance, with correlation of 0.926 for motion score index estimation and accuracy of 77.4\% for motion scoring. Cardiac-MOS also outperforms all existing work for binary abnormality detection. As the first automatic motion scoring solution, Cardiac-MOS demonstrates great potential in future clinical application.

Full Quantification of Left Ventricle via Deep Multitask Learning Network Respecting Intra- and Inter-Task Relatedness

Cardiac left ventricle (LV) quantification is among the most clinically important tasks for identification and diagnosis of cardiac diseases, yet still a challenge due to the high variability of cardiac structure and the complexity of temporal dynamics. Full quantification, i.e., to simultaneously quantify all LV indices including two areas (cavity and myocardium), six regional wall thicknesses (RWT), three LV dimensions, and one cardiac phase, is even more challenging since the uncertain relatedness intra and inter each type of indices may hinder the learning procedure from better convergence and generalization. In this paper, we propose a newly-designed multitask learning network (FullLVNet), which is constituted by a deep convolution neural network (CNN) for expressive feature embedding of cardiac structure; two followed parallel recurrent neural network (RNN) modules for temporal dynamic modeling; and four linear models for the final estimation. During the final estimation, both intra- and inter-task relatedness are modeled to enforce improvement of generalization: 1) respecting intra-task relatedness, group lasso is applied to each of the regression tasks for sparse and common feature selection and consistent prediction; 2) respecting inter-task relatedness, three phase-guided constraints are proposed to penalize violation of the temporal behavior of the obtained LV indices. Experiments on MR sequences of 145 subjects show that FullLVNet achieves high accurate prediction with our intra- and inter-task relatedness, leading to MAE of 190mm$^2$, 1.41mm, 2.68mm for average areas, RWT, dimensions and error rate of 10.4\% for the phase classification. This endows our method a great potential in comprehensive clinical assessment of global, regional and dynamic cardiac function.

Direct Estimation of Regional Wall Thicknesses via Residual Recurrent Neural Network

Accurate estimation of regional wall thicknesses (RWT) of left ventricular (LV) myocardium from cardiac MR sequences is of significant importance for identification and diagnosis of cardiac disease. Existing RWT estimation still relies on segmentation of LV myocardium, which requires strong prior information and user interaction. No work has been devoted into direct estimation of RWT from cardiac MR images due to the diverse shapes and structures for various subjects and cardiac diseases, as well as the complex regional deformation of LV myocardium during the systole and diastole phases of the cardiac cycle. In this paper, we present a newly proposed Residual Recurrent Neural Network (ResRNN) that fully leverages the spatial and temporal dynamics of LV myocardium to achieve accurate frame-wise RWT estimation. Our ResRNN comprises two paths: 1) a feed forward convolution neural network (CNN) for effective and robust CNN embedding learning of various cardiac images and preliminary estimation of RWT from each frame itself independently, and 2) a recurrent neural network (RNN) for further improving the estimation by modeling spatial and temporal dynamics of LV myocardium. For the RNN path, we design for cardiac sequences a Circle-RNN to eliminate the effect of null hidden input for the first time-step. Our ResRNN is capable of obtaining accurate estimation of cardiac RWT with Mean Absolute Error of 1.44mm (less than 1-pixel error) when validated on cardiac MR sequences of 145 subjects, evidencing its great potential in clinical cardiac function assessment.