Quantitative myocardial contrast echocardiography (MCE) has been proved to be valuable in detecting myocardial ischemia. During quantitative MCE analysis, myocardial segmentation is a critical step in determining accurate region of interests (ROIs). However, traditional myocardial segmentation mainly relies on manual tracing of myocardial contours, which is time-consuming and laborious. To solve this problem, we propose a fully automatic myocardial segmentation framework that can segment myocardial regions in MCE accurately without human intervention. A total of 100 patients’ MCE sequences were divided into a training set and a test set according to a 7: 3 proportion for analysis. We proposed a bi-directional training schema, which incorporated temporal information of forward and backward direction among frames in MCE sequences to ensure temporal consistency by combining convolutional neural network with recurrent neural network. Experiment results demonstrated that compared with a traditional segmentation model (U-net) and the model considering only forward temporal information (U-net + forward), our framework achieved the highest segmentation precision in Dice coefficient (U-net vs U-net + forward vs our framework: 0.78 ± 0.07 vs 0.79 ± 0.07 vs 0.81 ± 0.07, p < 0.01), Intersection over Union (0.65 ± 0.09 vs 0.66 ± 0.09 vs 0.68 ± 0.09, p < 0.01), and lowest Hausdorff Distance (32.68 ± 14.6 vs 28.69 ± 13.18 vs 27.59 ± 12.82 pixel point, p < 0.01). In the visual grading study, the performance of our framework was the best among these three models (52.47 ± 4.29 vs 54.53 ± 5.10 vs 57.30 ± 4.73, p < 0.01). A case report on a randomly selected subject for perfusion analysis showed that the perfusion parameters generated by using myocardial segmentation of our proposed framework were similar to that of the expert annotation. The proposed framework could generate more precise myocardial segmentation when compared with traditional methods. The perfusion parameters generated by these myocardial segmentations have a good similarity to that of manual annotation, suggesting that it has the potential to be utilized in routine clinical practice.

A deep learning approach with temporal consistency for automatic myocardial segmentation of quantitative myocardial contrast echocardiography

Deep neural networks have demonstrated their great potential in recent years, exceeding the performance of human experts in a wide range of applications. Due to their large sizes, however, compress...

Quantization of Deep Neural Networks for Accurate Edge Computing

Image denoising of low-dose CT images is a key problem in modern medical practice. Recently, several works adopted Convolutional Neural Network (CNN) to precisely capture the similarity between local features resulting in significant improvements. However, we discovered that the main drawback of existing works is the lack of non-local feature processing. On the other hand, currently, graph convolutional networks (GCN) have been widely used to process non-Euclidean geometry data considering both local and non-local features. Motivated by the property of GCN, in this paper, we propose an encoder-decoder-based graph convolutional network (ED-GCN) for CT image denoising. Particularly, we combine local convolutions and graph convolutions to process both local and non-local features. We collected seven CT volumes with Gaussian noise and Poisson noise in the experiment. Experimental results show that the proposed method outperforms existing CNN-based approaches significantly.

Ct Image Denoising With Encoder-Decoder Based Graph Convolutional Networks

Deep learning had already demonstrated its power in medical images, including denoising, classification, segmentation, etc. All these applications are proposed to automatically analyze medical images beforehand, which brings more information to radiologists during clinical assessment for accuracy improvement. Recently, many medical denoising methods had shown their significant artifact reduction result and noise removal both quantitatively and qualitatively. However, those existing methods are developed around human-vision, i.e., they are designed to minimize the noise effect that can be perceived by human eyes. In this paper, we introduce an application-guided denoising framework, which focuses on denoising for the following neural networks. In our experiments, we apply the proposed framework to different datasets, models, and use cases. Experimental results show that our proposed framework can achieve a better result than human-vision denoising network.

Do Noises Bother Human and Neural Networks In the Same Way? A Medical Image Analysis Perspective.

Multi-class segmentation of cardiac magnetic resonance (CMR) images seeks a separation of data into anatomical components with known structure and configuration. The most popular CNN-based methods are optimised using pixel wise loss functions, ignorant of the spatially extended features that characterise anatomy. Therefore, whilst sharing a high spatial overlap with the ground truth, inferred CNN-based segmentations can lack coherence, including spurious connected components, holes and voids. Such results are implausible, violating anticipated anatomical topology. In response, (single-class) persistent homology-based loss functions have been proposed to capture global anatomical features. Our work extends these approaches to the task of multi-class segmentation. Building an enriched topological description of all class labels and class label pairs, our loss functions make predictable and statistically significant improvements in segmentation topology using a CNN-based post-processing framework. We also present (and make available) a highly efficient implementation based on cubical complexes and parallel execution, enabling practical application within high resolution 3D data for the first time. We demonstrate our approach on 2D short axis and 3D whole heart CMR segmentation, advancing a detailed and faithful analysis of performance on two publicly available datasets.

A persistent homology-based topological loss for CNN-based multi-class segmentation of CMR

Congenital heart disease (CHD) is the most common type of birth defects, which occurs 1 in every 110 births in the United States. CHD usually comes with severe variations in heart structure and great artery connections that can be classified into many types. Thus highly specialized domain knowledge and time-consuming human process is needed to analyze the associated medical images. On the other hand, due to the complexity of CHD and the lack of dataset, little has been explored on the automatic diagnosis (classification) of CHDs. In this paper, we present ImageCHD, the first medical image dataset for CHD classification. ImageCHD contains 110 3D Computed Tomography (CT) images covering most types of CHD, which is of decent size compared with existing medical imaging datasets. Classification of CHDs requires the identification of large structural changes without any local tissue changes, with limited data. It is an example of a larger class of problems that are quite difficult for current machine-learning based vision methods to solve. To demonstrate this, we further present a baseline framework for automatic classification of CHD, based on a state-of-the-art CHD segmentation method. Experimental results show that the baseline framework can only achieve a classification accuracy of 82.0% under selective prediction scheme with 88.4% coverage, leaving big room for further improvement. We hope that ImageCHD can stimulate further research and lead to innovative and generic solutions that would have an impact in multiple domains. Our dataset is released to the public [1].

ImageCHD: A 3D Computed Tomography Image Dataset for Classification of Congenital Heart Disease

The invention discloses a congenital heart disease classification system based on a deep neural network and morphological similarity, and the system comprises an image segmentation module which is used for segmenting a medical image to obtain a first segmented image and a second segmented image; an atrioventricular and myocardial morphological characteristic analysis module, used for analyzing morphological characteristics of four atrioventricles and myocardium of the heart; a connection characteristic analysis module, used for analyzing connection relations between central atriums of the first segmentation image and the second segmentation image and a ventricle and between the ventricle and a great vessel; a blood vessel form extraction module, used for extracting a blood vessel form in the second segmented image; a template storage module, used for storing templates of all blood vessel forms of the congenital heart disease; a large blood vessel morphological characteristic analysis module, used for comparing the blood vessel morphology with the template; and a comprehensive analysis module, used for obtaining a classification result of the congenital heart disease according to apreset classification rule. The invention further provides a method for applying the classification system. According to the invention, congenital heart disease automatic classification can be carriedout on medical images, and the working efficiency of doctors is improved.

Congenital heart disease classification system and method based on deep neural network and morphological similarity

The invention discloses an image conversion method based on a multi-cycle generative adversarial network, and the method comprises the steps: obtaining an image which needs to be converted, and dividing the image into a first image type and a second image type which are different; constructing at least one intermediate image category according to the images of the first image category and the images of the second image category; constructing a generative adversarial network comprising a plurality of generators and a plurality of discriminators according to the first image category, the secondimage category and the intermediate image category; wherein the first image category correspondingly generates a first discriminator, the second image category correspondingly generates a second discriminator, and the intermediate image category correspondingly generates a third discriminator; and performing image conversion through the generative adversarial network. According to the invention, the image conversion capability can be improved so as to complete more complex and difficult conversion tasks.

Image conversion method and device based on multi-cycle generative adversarial network

Congenital heart disease (CHD) is the most common type of birth defect, which occurs 1 in every 110 births in the United States. CHD usually comes with severe variations in heart structure and great artery connections that can be classified into many types. Thus highly specialized domain knowledge and the time-consuming human process is needed to analyze the associated medical images. On the other hand, due to the complexity of CHD and the lack of dataset, little has been explored on the automatic diagnosis (classification) of CHDs. In this paper, we present ImageCHD, the first medical image dataset for CHD classification. ImageCHD contains 110 3D Computed Tomography (CT) images covering most types of CHD, which is of decent size Classification of CHDs requires the identification of large structural changes without any local tissue changes, with limited data. It is an example of a larger class of problems that are quite difficult for current machine-learning-based vision methods to solve. To demonstrate this, we further present a baseline framework for the automatic classification of CHD, based on a state-of-the-art CHD segmentation method. Experimental results show that the baseline framework can only achieve a classification accuracy of 82.0\% under a selective prediction scheme with 88.4\% coverage, leaving big room for further improvement. We hope that ImageCHD can stimulate further research and lead to innovative and generic solutions that would have an impact in multiple domains. Our dataset is released to the public compared with existing medical imaging datasets.

Cen Jianzheng

Papers

ImageCHD: A 3D Computed Tomography Image Dataset for Classification of Congenital Heart Disease

Congenital heart disease classification system and method based on deep neural network and morphological similarity

Image conversion method and device based on multi-cycle generative adversarial network

ImageCHD: A 3D Computed Tomography Image Dataset for Classification of Congenital Heart Disease