3D Image Segmentation Research Articles

ACCT is a fast and accessible automatic cell counting tool using machine learning for 2D image segmentation

Counting cells is a cornerstone of tracking disease progression in neuroscience. A common approach for this process is having trained researchers individually select and count cells within an image, which is not only difficult to standardize but also very time-consuming. While tools exist to automatically count cells in images, the accuracy and accessibility of such tools can be improved. Thus, we introduce a novel tool ACCT: Automatic Cell Counting with Trainable Weka Segmentation which allows for flexible automatic cell counting via object segmentation after user-driven training. ACCT is demonstrated with a comparative analysis of publicly available images of neurons and an in-house dataset of immunofluorescence-stained microglia cells. For comparison, both datasets were manually counted to demonstrate the applicability of ACCT as an accessible means to automatically quantify cells in a precise manner without the need for computing clusters or advanced data preparation.

An efficient memory reserving-and-fading strategy for vector quantization based 3D brain segmentation and tumor extraction using an unsupervised deep learning network.

Deep learning networks are state-of-the-art approaches for 3D brain image segmentation, and the radiological characteristics extracted from tumors are of great significance for clinical diagnosis, treatment planning, and treatment outcome evaluation. However, two problems have been the hindering factors in brain image segmentation techniques. One is that deep learning networks require large amounts of manually annotated data. Another issue is the computational efficiency of 3D deep learning networks. In this study, we propose a vector quantization (VQ)-based 3D segmentation method that employs a novel unsupervised 3D deep embedding clustering (3D-DEC) network and an efficiency memory reserving-and-fading strategy. The VQ-based 3D-DEC network is trained on volume data in an unsupervised manner to avoid manual data annotation. The memory reserving-and-fading strategy beefs up model efficiency greatly. The designed methodology makes deep learning-based model feasible for biomedical image segmentation. The experiment is divided into two parts. First, we extensively evaluate the effectiveness and robustness of the proposed model on two authoritative MRI brain tumor databases (i.e., IBSR and BrainWeb). Second, we validate the model using real 3D brain tumor data collected from our institute for clinical practice significance. Results show that our method (without data manual annotation) has superior accuracy ( Tanimoto coefficient on IBSR, 97.5% TP and 97.7% TN on BrainWeb, and 91% Dice, 88% sensitivity and 87% specificity on real brain data) and remarkable efficiency (speedup ratio is 18.72 on IBSR, 31.16 on BrainWeb, 31.00 on real brain data) compared to the state-of-the-art methods. The results show that our proposed model can address the lacks of manual annotations, and greatly increase computation speedup with competitive segmentation accuracy compared to other state-of-the-art 3D CNN models. Moreover, the proposed model can be used for tumor treatment follow-ups every 6months, providing critical details for surgical and postoperative treatment by correctly extracting numerical radiomic features of tumors.

MAGNET: A MODALITY-AGNOSTIC NETWORK FOR 3D MEDICAL IMAGE SEGMENTATION.

In this paper, we proposed MAGNET, a novel modality-agnostic network for 3D medical image segmentation. Different from existing learning methods, MAGNET is specifically designed to handle real medical situations where multiple modalities/sequences are available during model training, but fewer ones are available or used at time of clinical practice. Our results on multiple datasets show that MAGNET trained on multi-modality data has the unique ability to perform predictions using any subset of training imaging modalities. It outperforms individually trained uni-modality models while can further boost performance when more modalities are available at testing.

Characterizing browser-based medical imaging AI with serverless edge computing: towards addressing clinical data security constraints.

Artificial intelligence (AI) has been widely introduced to various medical imaging applications ranging from disease visualization to medical decision support. However, data privacy has become an essential concern in clinical practice of deploying the deep learning algorithms through cloud computing. The sensitivity of patient health information (PHI) commonly limits network transfer, installation of bespoke desktop software, and access to computing resources. Serverless edge-computing shed light on privacy preserved model distribution maintaining both high flexibility (as cloud computing) and security (as local deployment). In this paper, we propose a browser-based, cross-platform, and privacy preserved medical imaging AI deployment system working on consumer-level hardware via serverless edge-computing. Briefly we implement this system by deploying a 3D medical image segmentation model for computed tomography (CT) based lung cancer screening. We further curate tradeoffs in model complexity and data size by characterizing the speed, memory usage, and limitations across various operating systems and browsers. Our implementation achieves a deployment with (1) a 3D convolutional neural network (CNN) on CT volumes (256×256×256 resolution), (2) an average runtime of 80 seconds across Firefox v.102.0.1/Chrome v.103.0.5060.114/Microsoft Edge v.103.0.1264.44 and 210 seconds on Safari v.14.1.1, and (3) an average memory usage of 1.5 GB on Microsoft Windows laptops, Linux workstation, and Apple Mac laptops. In conclusion, this work presents a privacy-preserved solution for medical imaging AI applications that minimizes the risk of PHI exposure. We characterize the tools, architectures, and parameters of our framework to facilitate the translation of modern deep learning methods into routine clinical care.

3D Mesh classification and panoramic image segmentation using spherical vector networks with rotation-equivariant self-attention mechanism

Spherical signals exist in many applications such as planetary data, lidar scanning and digitization of 3D objects, so we need models that can effectively process spherical data. When the spherical data is simply projected onto a two-dimensional plane and then convolutional neural networks (CNNs) are used, the performance of the previous algorithms that exist in the literature is poor due to the distortion caused by the projection and the invalid translational equivariance. We propose a spherical vector network with rotation-equivariant self-attention mechanism for part-whole relationships learning to avoid a certain degree of distortion in this paper. Specifically, we take first the spherical convolutional network as the front-end network to obtain primary vectors, then we achieve the part-whole relationships between vectors through proposed rotation-equivariant self-attention mechanism to obtain advanced vectors which can represent the existence probability of the entity and orientations. Experimental results show that the proposed method combined with the front-end network improves the 3D mesh classification accuracy of the front-end network by 9% when the training set is not rotated and the test set is rotated arbitrarily under the rigid ModelNet40 dataset. Similarly, the 3D mesh classification accuracy of the front-end network improves by 12.2% under the non-rigid SHREC15 dataset. In addition, our method is compared with the recent method in the spherical image semantic segmentation task, achieving an improvement of 2.2% in mean pixel accuracy and 1.3% in mean intersection over union.

VLSM-Net: A Fusion Architecture for CT Image Segmentation

Region of interest (ROI) segmentation is a key step in computer-aided diagnosis (CAD). With the problems of blurred tissue edges and imprecise boundaries of ROI in medical images, it is hard to extract satisfactory ROIs from medical images. In order to overcome the shortcomings in segmentation from the V-Net model or the level set method (LSM), we propose in this paper a new image segmentation method, the VLSM-Net model, combining these two methods. Specifically, we first use the V-Net model to segment the ROIs, and set the segmentation result as the initial contour. It is then fed through the hybrid LSM for further fine segmentation. That is, the complete segmentation of the V-Net model can be obtained by successively combining the V-Net model and the hybrid LSM. The experimental results conducted in the public datasets LiTS and LUNA show that, compared with the V-Net model or LSM alone, our VLSM-Net model greatly improves the sensitivity, precision and dice coefficient values (DCV) in 3D image segmentation, thus validating our model’s effectiveness.

How can pore characteristics of soil aggregates from contrasting tillage systems affect their intrinsic permeability and hydraulic conductivity?

Studies on how the alteration in the soil pore architecture can influence the permeability of small soil aggregates are still scarce. These aggregates affect the water dynamics mainly at the intra-aggregate pore scale. Thus, investigating the intra-aggregate pore space properties becomes crucial for comprehending the matter and energy transported into the soil. This paper focuses on analyzing the permeability and hydraulic conductivity of small soil aggregates from contrasting tillage systems (no-tillage – NT, minimum tillage – MT, conventional tillage – CT, and secondary forest – F) through synchrotron-based X-ray Computed Tomography. The 3D image segmentation was performed using machine and deep learning strategies. Permeability was measured based on the image-based parametrization of the Kozeny-Carman equation, which is dependent on porosity, hydraulic radius, and tortuosity. The porosity, hydraulic radius, and pore connectivity were not affected by the soil tillage systems at the scale of analysis (aggregates of a few millimeters). Conventional tillage had higher pore tortuosity relative to NT, demonstrating less favorable fluid transport at the aggregate scale. However, only the aggregates under F had significantly higher permeability and hydraulic conductivity relative to NT, MT, and CT. Our results brought new insights into the effects of different tillage systems on the soil structure and their possible impact on water dynamics at the microscale.

Segmentation, tracking, and sub-cellular feature extraction in 3D time-lapse images

This paper presents a method for time-lapse 3D cell analysis. Specifically, we consider the problem of accurately localizing and quantitatively analyzing sub-cellular features, and for tracking individual cells from time-lapse 3D confocal cell image stacks. The heterogeneity of cells and the volume of multi-dimensional images presents a major challenge for fully automated analysis of morphogenesis and development of cells. This paper is motivated by the pavement cell growth process, and building a quantitative morphogenesis model. We propose a deep feature based segmentation method to accurately detect and label each cell region. An adjacency graph based method is used to extract sub-cellular features of the segmented cells. Finally, the robust graph based tracking algorithm using multiple cell features is proposed for associating cells at different time instances. We also demonstrate the generality of our tracking method on C. elegans fluorescent nuclei imagery. Extensive experiment results are provided and demonstrate the robustness of the proposed method. The code is available on GitHub and the method is available as a service through the BisQue portal.

PLN: Parasitic-Like Network for Barely Supervised Medical Image Segmentation.

It is known that annotations for 3D medical image segmentation tasks are laborious, time-consuming and expensive. Considering the similarities existing in inter-slice and inter-volume, we believe that the delineation way and the model architecture should be tightly coupled. In this paper, by introducing an extremely sparse annotation way of labeling only one slice per 3D image, we investigate a novel barely-supervised segmentation setting with only a few sparsely-labeled images along with a large amount of unlabeled images. To achieve this goal, we present a new parasitic-like network including a registration module (as host) and a semi-supervised segmentation module (as parasite) to deal with inter-slice label propagation and inter-volume segmentation prediction, respectively. Specifically, our parasitism mechanism effectively achieves the collaboration of these two modules through three stages of infection, development and eclosion, providing accurate pseudo-labels for training. Extensive results demonstrate that our framework is capable of achieving high performance on extremely sparse annotation tasks, e.g., we achieve Dice of 84.83% on LA dataset with only 16 labeled slices. The code is available at https: //github.com/ShumengLI/PLN.

Large-Kernel Attention for 3D Medical Image Segmentation

Automated segmentation of multiple organs and tumors from 3D medical images such as magnetic resonance imaging (MRI) and computed tomography (CT) scans using deep learning methods can aid in diagnosing and treating cancer. However, organs often overlap and are complexly connected, characterized by extensive anatomical variation and low contrast. In addition, the diversity of tumor shape, location, and appearance, coupled with the dominance of background voxels, makes accurate 3D medical image segmentation difficult. In this paper, a novel 3D large-kernel (LK) attention module is proposed to address these problems to achieve accurate multi-organ segmentation and tumor segmentation. The advantages of biologically inspired self-attention and convolution are combined in the proposed LK attention module, including local contextual information, long-range dependencies, and channel adaptation. The module also decomposes the LK convolution to optimize the computational cost and can be easily incorporated into CNNs such as U-Net. Comprehensive ablation experiments demonstrated the feasibility of convolutional decomposition and explored the most efficient and effective network design. Among them, the best Mid-type 3D LK attention-based U-Net network was evaluated on CT-ORG and BraTS 2020 datasets, achieving state-of-the-art segmentation performance when compared to avant-garde CNN and Transformer-based methods for medical image segmentation. The performance improvement due to the proposed 3D LK attention module was statistically validated.

Evolution of microstructure and CO2 diffusion coefficient of compacted recycled aggregates during carbonation investigated by X-ray tomography

Recycled aggregates compacted in road layers are made of a majority of concrete aggregates which can mineralize atmospheric CO2. Previous studies showed that carbonation modifies the microstructure of recycled concrete aggregates, but little information is available in the case of compacted recycled aggregates. Furthermore, carbonation kinetics depends on CO2 diffusion through the compacted material, which is linked its microstructure. This paper presents an investigation of the evolution of both microstructure and CO2 diffusion coefficient of compacted recycled aggregates during carbonation in two different environments by X-ray tomography. Segmented 3D images were analyzed to fully extract resolved pore structures. In addition, the CO2 diffusion coefficient variations were investigated by calculating the macroscopic diffusivity tensor in the segmented 3D images. We show that the numerical results are very similar to previous experimental results from gas diffusion tests and that the diffusivity coefficients can be estimated by using only the larger pore network (pore size > 5.5 µm). These findings suggest that nondestructive imaging approaches, such as micro-Computed Tomography (µCT), are a relevant alternative to experimental characterization methods.

Automated segmentation of standard scanning planes to measure biometric parameters in foetal ultrasound images – a survey

ABSTRACT Accurate foetal ultrasound (US) image segmentation facilitates advanced obstetric health care by enabling remote monitoring of expectant mothers. However, foetal US image segmentation is challenging due to distortions, motion artefacts, various imaging conditions and presence of maternal anatomy. Recent research work has proposed many methods towards increasing the accuracy of foetal US image segmentation. This paper reviews 2D and 3D foetal US image segmentation methods under four main categories; deep learning-based method, machine learning-based methods, active contour-based methods and thresholding-based methods. Each of these methods are discussed highlighting their advantages, limitations and potential in contributing to further development. In addition, the paper highlights possible prospects that would streamline the future research work.

3D Quantitative Mineral Characterization of Particles Using X-ray Computed Tomography

A new method to measure and quantify the 3D mineralogical composition of particulate materials using X-ray computed micro-tomography (CT) is presented. The new method is part of a workflow designed to standardize the analysis of particles based on their microstructures without the need to segment the individual classes or grains. Classification follows a decision tree with criteria derived from particle histogram parameters that are specific to each microstructure, which in turn can be identified by 2D-based automated quantitative mineralogy. The quantification of mineral abundances is implemented at the particle level according to the complexity of the particle by taking into consideration the partial volume effect at interphases. The new method was tested on two samples with different particle size distributions from a carbonate rock containing various microstructures and phases. The method allowed differentiation and quantification of more mineral classes than traditional 3D image segmentation that uses only the grey-scale for mineral classification. Nevertheless, due to lower spatial resolution and lack of chemical information, not all phases identified in 2D could be distinguished. However, quantification of the mineral classes that could be distinguished was more representative than their 2D quantification, especially for coarser particle sizes and for minor phases. Therefore, the new 3D method shows great potential as a complement to 2D-based methods and as an alternative to traditional phase segmentation analysis of 3D images. Particle-based quantification of mineralogical and 3D geometrical properties of particles opens new applications in the raw materials and particle processing industries.

3D Medical image segmentation using parallel transformers

Most recent 3D medical image segmentation methods adopt convolutional neural networks (CNNs) that rely on deep feature representation and achieve adequate performance. However, due to the convolutional architectures having limited receptive fields, they cannot explicitly model the long-range dependencies in the medical image. Recently, Transformer can benefit from global dependencies using self-attention mechanisms and learn highly expressive representations. Some works were designed based on the Transformers, but the existing Transformers suffer from extreme computational and memories, and they cannot take full advantage of the powerful feature representations in 3D medical image segmentation. In this paper, we aim to connect the different resolution streams in parallel and propose a novel network, named Transformer based High Resolution Network (TransHRNet), with an Effective Transformer (EffTrans) block, which has sufficient feature representation even at high feature resolutions. Given a 3D image, the encoder first utilizes CNN to extract the feature representations to capture the local information, and then the different feature maps are reshaped elaborately for tokens that are fed into each Transformer stream in parallel to learn the global information and repeatedly exchange the information across streams. Unfortunately, the proposed framework based on the standard Transformer needs a huge amount of computation, thus we introduce a deep and effective Transformer to deliver better performance with fewer parameters. The proposed TransHRNet is evaluated on the Multi-Atlas Labeling Beyond the Cranial Vault (BCV) dataset that consists of 11 major human organs and the Medical Segmentation Decathlon (MSD) dataset for brain tumor and spleen segmentation tasks. Experimental results show that it performs better than the convolutional and other related Transformer-based methods on the 3D multi-organ segmentation tasks. Code is available at https://github.com/duweidai/TransHRNet.

Surface-GCN: Learning interaction experience for organ segmentation in 3D medical images.

Accurate segmentation of organs has a great significance for clinical diagnosis, but it is still hard work due to the obscure imaging boundaries caused by tissue adhesion on medical images. Based on the image continuity in medical image volumes, segmentation on these slices could be inferred from adjacent slices with a clear organ boundary. Radiologists can delineate a clear organ boundary by observing adjacent slices. Inspired by the radiologists' delineating procedure, we design an organ segmentation model based on boundary information of adjacent slices and a human-machine interactive learning strategy to introduce clinicalexperience. We propose an interactive organ segmentation method for medical image volume based on Graph Convolution Network (GCN) called Surface-GCN. First, we propose a Surface Feature Extraction Network (SFE-Net) to capture surface features of a target organ, and supervise it by a Mini-batch Adaptive Surface Matching (MBASM) module. Then, to predict organ boundaries precisely, we design an automatic segmentation module based on a Surface Convolution Unit (SCU), which propagates information on organ surfaces to refine the generated boundaries. In addition, an interactive segmentation module is proposed to learn radiologists' experience of interactive corrections on organ surfaces to reduce interactionclicks. We evaluate the proposed method on one prostate MR image dataset and two abdominal multi-organ CT datasets. The experimental results show that our method outperforms other state-of-the-artmethods. For prostate segmentation, the proposed method conducts a DSC score of 94.49% on PROMISE12 test dataset. For abdominal multi-organ segmentation, the proposed method achieves DSC scores of 95, 91, 95, and 88% for the left kidney, gallbladder, spleen, and esophagus, respectively. For interactive segmentation, the proposed method reduces 5-10 interaction clicks to reach the same accuracy. To overcome the medical organ segmentation challenge, we propose a Graph Convolutional Network called Surface-GCN by imitating radiologist interactions and learning clinical experience. On single and multiple organ segmentation tasks, the proposed method could obtain more accurate segmentation boundaries compared with other state-of-the-artmethods.

Illuminating dynamic structural states in hiPSC through endogenous gene tagging.

The Allen Institute for Cell Science aims to understand the principles by which human induced pluripotent stem cells (hiPSCs) establish and maintain robust dynamic localization of cellular structures, and how they transition between states during differentiation and disease. As a first step [JG1], we have created a collection of endogenously tagged hiPSC lines to illuminate cell organization. The tagged proteins include notable membrane-bound and membrane-less cellular organelles, signaling complexes, phase transition markers, transcription factors, and structural markers specific to differentiated cells. To date, the Allen Cell Collection consists of 54 single- or dual-edited lines that have undergone extensive quality control testing to ensure genomic, cell biological, and stem cell integrity. Here, we highlight our gene-editing and quality control workflow for mono- and biallelic editing of expressed or silent genes that are expressed specifically during differentiation and function as reporters of cellular state. Furthermore, we underscore our efforts to increase efficiency of the current workflow by utilizing Adeno-Associated Virus (AAV) to deliver donor DNAs and multiplexing transfection strategies for gene tagging at multiple loci. Our most recently released lines are CTCF-mEGFP and PCNA-mEGFP, for dynamic visualization of chromatin domains and replication foci, G3BP1-mEGFP to label stress granules, mono- and bi-allelic mEGFP-E-cadherin to visualize adherens junctions and DCP1a-mEGFP for P-bodies. We now have gene [RG2] [JG3]-edited cell populations for N- and VE-cadherin ready for single cell sorting. Our overarching goal is to make these tools openly available to the cell biology community to help accelerate biomedical research. In addition to our cell lines, the donor plasmids, segmented 3D images of about 200,000 cells from our lines, image analysis, visualization tools, integrated cell models and biological findings are all openly available to the research community (www.allencell.org).

A two-stage CNN method for MRI image segmentation of prostate with lesion

Prostate magnetic resonance imaging (MRI) is widely used in the diagnosis of prostate cancer and other prostate diseases. The automatic segmentation of images from prostate MRI plays an important role in the auxiliary diagnosis of prostate diseases. Currently, there are two commonly used methods for automatic segmentation of prostate MRI, which are 2D image segmentation and 3D image segmentation. In this paper, a two-stage CNN method for MRI image segmentation of prostate with lesion is proposed. At the first stage, we used a CNN model incorporating the Squeeze-Excitation module to discriminate whether the image contains prostate or not. At the second stage, we proposed a Residual-Attention U-Net for segmentation of images containing prostate. Eventually, the 3D prostate MRI segmentation results are obtained and fully automated segmentation is accomplished. We evaluated our proposed method and other common 2D and 3D segmentation methods on the test dataset and compared their results based on Dice Similarity Coefficient (DSC) value. Our method performed the best and achieved the DSC metric value of 0.860.

Automated multi-modal Transformer network (AMTNet) for 3D medical images segmentation

Objective. Over the past years, convolutional neural networks based methods have dominated the field of medical image segmentation. But the main drawback of these methods is that they have difficulty representing long-range dependencies. Recently, the Transformer has demonstrated super performance in computer vision and has also been successfully applied to medical image segmentation because of the self-attention mechanism and long-range dependencies encoding on images. To the best of our knowledge, only a few works focus on cross-modalities of image segmentation using the Transformer. Hence, the main objective of this study was to design, propose and validate a deep learning method to extend the application of Transformer to multi-modality medical image segmentation. Approach. This paper proposes a novel automated multi-modal Transformer network termed AMTNet for 3D medical image segmentation. Especially, the network is a well-modeled U-shaped network architecture where many effective and significant changes have been made in the feature encoding, fusion, and decoding parts. The encoding part comprises 3D embedding, 3D multi-modal Transformer, and 3D Co-learn down-sampling blocks. Symmetrically, the 3D Transformer block, upsampling block, and 3D-expanding blocks are included in the decoding part. In addition, a Transformer-based adaptive channel interleaved Transformer feature fusion module is designed to fully fuse features of different modalities. Main results. We provide a comprehensive experimental analysis of the Prostate and BraTS2021 datasets. The results show that our method achieves an average DSC of 0.907 and 0.851 (0.734 for ET, 0.895 for TC, and 0.924 for WT) on these two datasets, respectively. These values show that AMTNet yielded significant improvements over the state-of-the-art segmentation networks. Significance. The proposed 3D segmentation network exploits complementary features of different modalities during the feature extraction process at multiple scales to increase the 3D feature representations and improve the segmentation efficiency. This powerful network enriches the research of the Transformer to multi-modal medical image segmentation.

Topological structure and global features enhanced graph reasoning model for non-small cell lung cancer segmentation from CT

Objective. Accurate and automated segmentation of lung tumors from computed tomography (CT) images is critical yet challenging. Lung tumors are of various sizes and locations and have indistinct boundaries adjacent to other normal tissues. Approach. We propose a new segmentation model that can integrate the topological structure and global features of image region nodes to address the challenges. Firstly, we construct a weighted graph with image region nodes. The graph topology reflects the complex spatial relationships among these nodes, and each node has its specific attributes. Secondly, we propose a node-wise topological feature learning module based on a new graph convolutional autoencoder (GCA). Meanwhile, a node information supplementation (GNIS) module is established by integrating specific features of each node extracted by a convolutional neural network (CNN) into each encoding layer of GCA. Afterwards, we construct a global feature extraction model based on multi-layer perceptron (MLP) to encode the features learnt from all the image region nodes which are crucial complementary information for tumor segmentation. Main results. Ablation study results over the public lung tumor segmentation dataset demonstrate the contributions of our major technical innovations. Compared with other segmentation methods, our new model improves the segmentation performance and has generalization ability on different 3D image segmentation backbones. Our model achieved Dice of 0.7827, IoU of 0.6981, and HD of 32.1743 mm on the public dataset 2018 Medical Segmentation Decathlon challenge, and Dice of 0.7004, IoU of 0.5704 and HD of 64.4661 mm on lung tumor dataset from Shandong Cancer Hospital. Significance. The novel model improves automated lung tumor segmentation performance especially the challenging and complex cases using topological structure and global features of image region nodes. It is of great potential to apply the model to other CT segmentation tasks.

Semi-supervised dual-task cross-consistency constraint network for 3D cardiac MRI segmentation

目的 针对人体组织器官及病灶区域的3维图像分割是计算机辅助医疗诊断的重要前提，是医学影像3维可视化的重要技术基础。深度学习方法在医学图像分割任务中的成功通常取决于大量有标注数据。半监督学习利用未标注数据容易获取的优点，在模型训练过程中使用少量标注数据和大量未标注数据进行学习，缓解了数据标注昂贵耗时的问题，在医学图像分割中受到了广泛关注。为更好地利用无标注数据，提升医学图像分割效果，提出一种新的一致性正则方法用于半监督3维医学图像分割。方法 模型以V-Net为基础架构，通过扩展网络结构，在均带有分割任务及回归任务属性的双任务主副解码器之间添加了用于正则化约束的交叉损失，构建了具有形状感知的基于双任务的交叉一致性正则网络SACC-Net （shape-aware cross-consistency regular network based on dual tasks），实现将数据层面和模型层面的扰动融合进多任务机制的一致性正则方法，使模型能够更好地利用未标注数据的有效先验信息，并且具有更好的泛化性能。结果 在MICCAI 2018（Medical Image Computing and Computer Assisted Intervention Society）心房分割挑战赛公布的数据集中的3维左心房核磁共振成像上验证本文算法，在仅使用训练集中10%的有标注数据的实验组中，Dice系数、Jaccard指数、HD（Hausdorff distance）距离和平均对称表面距离分别为88.01%、78.89%、8.19和2.09。在另一组仅使用20%的有标注数据的实验中Dice系数、Jaccard指数、HD距离和平均对称表面距离分别达到90.11%、82.11%、6.57和1.78。结论 本文提出的半监督分割模型性能显著，综合了数据、模型和任务层面的一致性正则约束，与其他半监督方法相比分割效果更好且具有更佳的泛化性能。;Objective Three-dimensional（3D）image segmentation of human tissues，organs and lesion areas is projected for computer-aided diagnosis and medical images-related 3D visualization. Thanks to the emerging deep learning technique，fully-supervised network models have been developing intensively in relevant to medical image segmentation tasks. However，it is challenged for a large amount of annotated data and 3D image segmentation data-labeled is costly and inefficient. Semi-supervised learning is focused on a small size of data-labeled and sufficient data-unlabeled in terms of easy acquisition of unlabeled data，which can alleviate the cost and time-consuming problem of data labeling. Our research is focused on new consistency regular method for semi-supervised 3D medical image segmentation model. To improve the medical image segmentation effect，our model can use unlabeled data through the fusion of different consistency methods. Method The network model is demonstrated on the V-Net，which can remove the residual structure of encoding and decoding. To get efficient features of unlabeled data，the proposed shape-aware cross-consistency regular network is introduced via the V-Net network structure extension on the basis of an encoder and two independent decoders-involved dual tasks （shape-aware cross-consistency regular network based on dual tasks（SACC-Net）），which are divided into a main decoder a and an auxiliary decoder b. The output of encoder-shared is transmitted to the two decoders after noise disturbance. At the same time，the two decoders can output the prediction results after each iteration. To increase the generalization and anti-noise ability of the model，it can minimize the difference between the two parts of the results during the training process. Additionally，the proportion of labeled samples in the training samples is extremely small because the feature distributions between the pre-processed medical image samples are relatively similar. To improve the learning ability of the model to segmented samples further，geometric prior information constraints are melted into the segmentation target. A shape-aware regression layer is added at the end of each decoder as well. During the training phase，each decoder can output two parts of the prediction results at the same time. That is，the total output of each iteration-after network consists of four parts. It can be used to decode the segmentation map SA and the signed distance map output by the decoder A，and the segmentation map SB and the signed distance map output by the decoder B through the dual-task consistency of each decoding part. To enhance the model’s ability and learn the effective features of segmentation targets to a greater extent， constraints and cross constraints can be used to realize a consistent regular method that combines data-level and modellevel disturbances with multi-task mechanisms，and make better use of unlabeled data. Result Our algorithm is validated on the MRI data set published in Atrium Segmentation Challenge held by MICCAI（Medical Image Computing and Computer Assisted Intervention Society）in 2018. The experiment is divided into two test groups based on the amounts proportion of labeled data. In the training set，10% annotated data is used only in the experimental group，the Dice coefficient， Jaccard index，HD （Hausdorff distance） distance，and average symmetric surface distance is reached to 88. 01%， 78. 89%，8. 19，and 2. 09 of each. In the other group，20% annotated data of the experiments are used only. The median Dice coefficient，Jaccard index，HD distance and average symmetric surface distance can be reached to 90. 14%， 82. 11%，6. 57，and 1. 78 each as well. Furthermore，In respect of the shape perception method using the level set function for regression tasks，the Dice evaluation index can be improved by 0. 69 and 0. 60 in comparison with shape-aware semi-supervised net（SASS Net）in 10% and 20% of the marked training results. Each improvement is reached to 1. 44% and 0. 72% in terms of comparative results of dual-task consistency（DTC）trained with 10% and 20% labeled data. Conclusion The semi-supervised segmentation model（SACC-Net）is illustrated for the criteria optimization for both of the region and boundary-based segmentation，which can incorporate the level-consistency of its data，model and task. The constrained method has its potential segmentation effect and generalization performance for semi-supervised methods.