Robust and efficient abdominal CT segmentation using shape constrained multi-scale attention network

PURPOSE

Although many deep learning-based abdominal multi-organ segmentation networks have been proposed, the various intensity distributions and organ shapes of the CT images from multi-center, multi-phase with various diseases introduce new challenges for robust abdominal CT segmentation. To achieve robust and efficient abdominal multi-organ segmentation, a new two-stage method is presented in this study.

METHODS

A binary segmentation network is used for coarse localization, followed by a multi-scale attention network for the fine segmentation of liver, kidney, spleen, and pancreas. To constrain the organ shapes produced by the fine segmentation network, an additional network is pre-trained to learn the shape features of the organs with serious diseases and then employed to constrain the training of the fine segmentation network.

RESULTS

The performance of the presented segmentation method was extensively evaluated on the multi-center data set from the Fast and Low GPU Memory Abdominal oRgan sEgmentation (FLARE) challenge, which was held in conjunction with International Conference on Medical Image Computing and Computer Assisted Intervention (MICCAI) 2021. Dice Similarity Coefficient (DSC) and Normalized Surface Dice (NSD) were calculated to quantitatively evaluate the segmentation accuracy and efficiency. An average DSC and NSD of 83.7% and 64.4% were achieved, and our method finally won the second place among more than 90 participating teams.

CONCLUSIONS

The evaluation results on the public challenge demonstrate that our method shows promising performance in robustness and efficiency, which may promote the clinical application of the automatic abdominal multi-organ segmentation.

Ageing and degeneration analysis using ageing-related dynamic attention on lateral cephalometric radiographs

With the increase of the ageing in the world’s population, the ageing and degeneration studies of physiological characteristics in human skin, bones, and muscles become important topics. Research on the ageing of bones, especially the skull, are paid much attention in recent years. In this study, a novel deep learning method representing the ageing-related dynamic attention (ARDA) is proposed. The proposed method can quantitatively display the ageing salience of the bones and their change patterns with age on lateral cephalometric radiographs images (LCR) images containing the craniofacial and cervical spine. An age estimation-based deep learning model based on 14142 LCR images from 4 to 40 years old individuals is trained to extract ageing-related features, and based on these features the ageing salience maps are generated by the Grad-CAM method. All ageing salience maps with the same age are merged as an ARDA map corresponding to that age. Ageing salience maps show that ARDA is mainly concentrated in three regions in LCR images: the teeth, craniofacial, and cervical spine regions. Furthermore, the dynamic distribution of ARDA at different ages and instances in LCR images is quantitatively analyzed. The experimental results on 3014 cases show that ARDA can accurately reflect the development and degeneration patterns in LCR images.

Convolutional Autoencoding and Gaussian Mixture Clustering for Unsupervised Beat-to-Beat Heart Rate Estimation of Electrocardiograms from Wearable Sensors

Heart rate is one of the most important diagnostic bases for cardiovascular disease. This paper introduces a deep autoencoding strategy into feature extraction of electrocardiogram (ECG) signals, and proposes a beat-to-beat heart rate estimation method based on convolution autoencoding and Gaussian mixture clustering. The high-level heartbeat features were first extracted in an unsupervised manner by training the convolutional autoencoder network, and then the adaptive Gaussian mixture clustering was applied to detect the heartbeat locations from the extracted features, and calculated the beat-to-beat heart rate. Compared with the existing heartbeat classification/detection methods, the proposed unsupervised feature learning and heartbeat clustering method does not rely on accurate labeling of each heartbeat location, which could save a lot of time and effort in human annotations. Experimental results demonstrate that the proposed method maintains better accuracy and generalization ability compared with the existing ECG heart rate estimation methods and could be a robust long-time heart rate monitoring solution for wearable ECG devices.

Automated Segmentation Method for Low Field 3D Stomach MRI Using Transferred Learning Image Enhancement Network

Accurate segmentation of abdominal organs has always been a difficult problem, especially for organs with cavities. And MRI-guided radiotherapy is particularly attractive for abdominal targets compared with low CT contrast. But in the limit of radiotherapy environment, only low field MRI segmentation can be used for stomach location, tracking, and treatment planning. In clinical applications, the existing 3D segmentation network model is trained by the low field MRI, and the segmentation result cannot be used in radiotherapy plan since the bad segmentation performance. Another way is that historical high field intensity MR images are directly used for data expansion to network learning; there will be a domain shift problem. How to use different domain images to improve the segmentation accuracy of deep neural network? A 3D low field MRI stomach segmentation method based on transfer learning image enhancement is proposed in this paper. In this method, Cycle Generative Adversarial Network (CycleGAN) is used to construct and learn the mapping relationship between high and low field intensity MRI and to overcome domain shift. Then, the image generated by the high field intensity MRI through the CycleGAN network is with transferred information as the extended data. The low field MRI combines these extended datasets to form the training data for training the 3D Res-Unet segmentation network. Furthermore, the convolution layer, batch normalization layer, and Relu layer together were replaced with a residual module to relieve the gradient disappearance of the neural network. The experimental results show that the Dice coefficient is 2.5 percent better than the baseline method. The over segmentation and under segmentation are reduced by 0.7 and 5.5 percent, respectively. And the sensitivity is improved by 6.4 percent.

Automated Classification and Segmentation in Colorectal Images Based on Self-Paced Transfer Network

Colorectal imaging improves on diagnosis of colorectal diseases by providing colorectal images. Manual diagnosis of colorectal disease is labor-intensive and time-consuming. In this paper, we present a method for automatic colorectal disease classification and segmentation. Because of label unbalanced and difficult colorectal data, the classification based on self-paced transfer VGG network (STVGG) is proposed. ImageNet pretraining network parameters are transferred to VGG network with training colorectal data to acquire good initial network performance. And self-paced learning is used to optimize the network so that the classification performance of label unbalanced and difficult samples is improved. In order to assist the colonoscopist to accurately determine whether the polyp needs surgical resection, feature of trained STVGG model is shared to Unet segmentation network as the encoder part and to avoid repeat learning of polyp segmentation model. The experimental results on 3061 colorectal images illustrated that the proposed method obtained higher classification accuracy (96%) and segmentation performance compared with a few other methods. The polyp can be segmented accurately from around tissues by the proposed method. The segmentation results underpin the potential of deep learning methods for assisting colonoscopist in identifying polyps and enabling timely resection of these polyps at an early stage.

Semi-Supervised Deep Metric Learning Networks for Classification of Polarimetric SAR Data

Recently, classification methods based on deep learning have attained sound results for the classification of Polarimetric synthetic aperture radar (PolSAR) data. However, they generally require a great deal of labeled data to train their models, which limits their potential real-world applications. This paper proposes a novel semi-supervised deep metric learning network (SSDMLN) for feature learning and classification of PolSAR data. Inspired by distance metric learning, we construct a network, which transforms the linear mapping of metric learning into the non-linear projection in the layer-by-layer learning. With the prior knowledge of the sample categories, the network also learns a distance metric under which all pairs of similarly labeled samples are closer and dissimilar samples have larger relative distances. Moreover, we introduce a new manifold regularization to reduce the distance between neighboring samples since they are more likely to be homogeneous. The categorizing is achieved by using a simple classifier. Several experiments on both synthetic and real-world PolSAR data from different sensors are conducted and they demonstrate the effectiveness of SSDMLN with limited labeled samples, and SSDMLN is superior to state-of-the-art methods.

Large-Deformation Image Registration of CT-TEE for Surgical Navigation of Congenital Heart Disease

The surgical treatment of congenital heart disease requires navigational assistance with transesophageal echocardiography (TEE); however, TEE images are often difficult to interpret and provide very limited anatomical information. Registering preoperative CT images to intraoperative TEE images provides surgeons with richer and more useful anatomical information. Yet, CT and TEE images differ substantially in terms of scale and geometry. In the present research, we propose a novel method for the registration of CT and TEE images for navigation during surgical repair of large defects in patients with congenital heart disease. Valve data was used for the coarse registration to determine the basic location. This was followed by the use of an enhanced probability model map to overcome gray-level differences between the two imaging modalities. Finally, the rapid optimization of mutual information was achieved by migrating parameters. This method was tested on a dataset of 240 images from 12 infant, children (≤ 3 years old), and adult patients with congenital heart disease. Compared to the “bronze standard” registration, the proposed method was more accurate with an average Dice coefficient of 0.91 and an average root mean square of target registration error of 1.2655 mm.