Segmentation Results Research Articles

MFH‐Net: A Hybrid CNN‐Transformer Network Based Multi‐Scale Fusion for Medical Image Segmentation

ABSTRACTIn recent years, U‐Net and its variants have gained widespread use in medical image segmentation. One key aspect of U‐Net's design is the skip connection, facilitating the retention of detailed information and leading to finer segmentation results. However, existing research often concentrates on enhancing either the encoder or decoder, neglecting the semantic gap between them, and resulting in suboptimal model performance. In response, we introduce Multi‐Scale Fusion module aimed at enhancing the original skip connections and addressing the semantic gap. Our approach fully incorporates the correlation between outputs from adjacent encoder layers and facilitates bidirectional information exchange across multiple layers. Additionally, we introduce Channel Relation Perception module to guide the fused multi‐scale information for efficient connection with decoder features. These two modules collectively bridge the semantic gap by capturing spatial and channel dependencies in the features, contributing to accurate medical image segmentation. Building upon these innovations, we propose a novel network called MFH‐Net. On three publicly available datasets, ISIC2016, ISIC2017, and Kvasir‐SEG, we perform a comprehensive evaluation of the network. The experimental results show that MFH‐Net exhibits higher segmentation accuracy in comparison with other competing methods. Importantly, the modules we have devised can be seamlessly incorporated into various networks, such as U‐Net and its variants, offering a potential avenue for further improving model performance.

HyFormer: a hybrid transformer-CNN architecture for retinal OCT image segmentation

Optical coherence tomography (OCT) has become the leading imaging technique in diagnosing and treatment planning for retinal diseases. Retinal OCT image segmentation involves extracting lesions and/or tissue structures to aid in the decisions of ophthalmologists, and multi-class segmentation is commonly needed. As the target regions often spread widely inside the retina, and the intensities and locations of different categories can be close, good segmentation networks must possess both global modeling capabilities and the ability to capture fine details. To address the challenge in capturing both global and local features simultaneously, we propose HyFormer, an efficient, lightweight, and robust hybrid network architecture. The proposed architecture features parallel Transformer and convolutional encoders for independent feature capture. A multi-scale gated attention block and a group positional embedding block are introduced within the Transformer encoder to enhance feature extraction. Feature integration is achieved in the decoder composed of the proposed three-path fusion modules. A class activation map-based cross-entropy loss function is also proposed to improve segmentation results. Evaluations are performed on a private dataset with myopic traction maculopathy lesions and the public AROI dataset for retinal layer and lesion segmentation with age-related degeneration. The results demonstrate HyFormer's superior segmentation performance and robustness compared to existing methods, showing promise for accurate and efficient OCT image segmentation. .

Fasteners quantitative detection and lightweight deployment based on improved YOLOv8.

Currently, research on on-board real-time quantitative detection of rail fasteners is few. Therefore, this paper proposes and validates an improved YOLOv8 based method for quantitative detection of rail fasteners, leveraging the capabilities of edge miniaturized artificial intelligence (AI) computing devices. First, the lightweight MobileNetV3 is employed as the backbone network for our model to increase detection speed, and the SA attention mechanism is integrated at the end of the backbone network to enhance the feature of the fasteners. Then the deformable convolution is introduced to reconstruct the bottleneck structure of the neck network, which can segment fasteners without compromising accuracy. Subsequently, the optimized network model is utilized on a Jetson AGX Xavier edge AI computing device by the TensorRT acceleration method. Segmentation results are then extracted at the pixel level for quantitative analysis of fastener breakage degree and deflection angle, so as to correct the detection results. Experimental results show that the size of the improved lightweight network volume is reduced by 28% compared to the original YOLOv8 model, and the frame rate on the edge AI computing device is lifted by 71.87%, i.e., 55 f/s. Furthermore, the model is refined based on quantitative analysis results, resulting in an mAP0.5 of 97.0%, and real-time quantitative detection of rail fasteners is realized.

On-Line Measurement of Tracking Poses of Heliostats in Concentrated Solar Power Plants.

The tracking pose of heliostats directly affects the stability and working efficiency of concentrated solar power (CSP) plants. Due to occlusion, over-exposure, and uneven illumination caused by mirror reflection, traditional image processing algorithms showed poor performances on the detection and segmentation of heliostats, which impede vision-based 3D measurement of tracking poses of heliostats. To tackle this issue, object detection using deep learning neural networks are exploited. An improved neural network based on YOLO-v5 framework has been designed to solve the on-line detection problem of heliostats. The model achieves a recognition accuracy of 99.7% for the test set, outperforming traditional methods significantly. Based on segmented results, the corner points of each heliostat are found out using Hough Transform and line intersection methods. The 3D poses of each heliostat are then solved out based on the image coordinates of specific feature points and the camera model. Experimental and field test results demonstrate the feasibility of this hybrid approach, which provides a low-cost solution for the monitoring and measurement of tracking poses of the heliostats in CSP.

Modelling a neural network for analysing the results of segmentation of satellite images

Modelling a neural network for analysing the results of segmentation of satellite images

Large-scale inventory in natural forests with mobile LiDAR point clouds

Large-scale forest inventory at the individual tree level is critical for natural resource management decision making. Terrestrial Laser Scanning (TLS) has been used for individual tree level inventory at plot scale However, due to the inflexibility of TLS and the complex scene of natural forests, it is still challenging to localize and measure every tree at large scale. In this paper, we present a framework to conduct large-scale natural forest inventory at the individual tree level by taking advantage of deep learning models and Mobile Laser Scanning (MLS) systems. First, a deep learning model, ForestSPG, was developed to perform large-scale semantic segmentation on MLS LiDAR data in natural forests. Then, the forest segmentation results were used for individual stem mapping. Finally, Diameter at Breast Height (DBH) was measured for each individual stem. Two natural forests mapped with backpack and Unmanned Aerial Vehicle (UAV) LiDAR systems were tested. The results showed that the proposed ForestSPG is able to segment large-scale forest LiDAR data into multiple ecologically meaningful classes. The proposed framework was able to localize and measure all 5838 stems at individual tree level in a 20 ha natural forest in less than 20 min using UAV LiDAR. DBH measurement results on trees’ DBH larger than 38.1 cm (15 in) showed that backpack LiDAR was able to achieve 1.82 cm of Root Mean Square Error (RMSE) and UAV LiDAR was able to achieve 3.13 cm of RMSE. The proposed framework can not only segment complex forest components with LiDAR data from different platforms but also demonstrate good performance on stem mapping and DBH measurement. Our research provides and automatic and scalable solution for large-scale natural forest inventory at individual tree level, which can be the basis for large-scale estimation of wood volume and biomass.

Attention-guided hierarchical fusion U-Net for uncertainty-driven medical image segmentation

Small inaccuracies in the system components or artificial intelligence (AI) models for medical imaging could have significant consequences leading to life hazards. To mitigate those risks, one must consider the precision of the image analysis outcomes (e.g., image segmentation), along with the confidence in the underlying model predictions. U-shaped architectures, based on the convolutional encoder–decoder, have established themselves as a critical component of many AI-enabled diagnostic imaging systems. However, most of the existing methods focus on producing accurate diagnostic predictions without assessing the uncertainty associated with such predictions or the introduced techniques. Uncertainty maps highlight areas in the predicted segmented results, where the model is uncertain or less confident. This could lead radiologists to pay more attention to ensuring patient safety and pave the way for trustworthy AI applications. In this paper, we therefore propose the Attention-guided Hierarchical Fusion U-Net (named AHF-U-Net) for medical image segmentation. We then introduce the uncertainty-aware version of it called UA-AHF-U-Net which provides the uncertainty map alongside the predicted segmentation map. The network is designed by integrating the Encoder Attention Fusion module (EAF) and the Decoder Attention Fusion module (DAF) on the encoder and decoder sides of the U-Net architecture, respectively. The EAF and DAF modules utilize spatial and channel attention to capture relevant spatial information and indicate which channels are appropriate for a given image. Furthermore, an enhanced skip connection is introduced and named the Hierarchical Attention-Enhanced (HAE) skip connection. We evaluated the efficiency of our model by comparing it with eleven well-established methods for three popular medical image segmentation datasets consisting of coarse-grained images with unclear boundaries. Based on the quantitative and qualitative results, the proposed method ranks first in two datasets and second in a third. The code can be accessed at: https://github.com/AfsanaAhmedMunia/AHF-Fusion-U-Net

Uncertainty guided semi-supervised few-shot segmentation with prototype level fusion

Few-Shot Semantic Segmentation (FSS) aims to tackle the challenge of segmenting novel categories with limited annotated data. However, given the diversity among support-query pairs, transferring meta-knowledge to unseen categories poses a significant challenge, particularly in scenarios featuring substantial intra-class variance within an episode task. To alleviate this issue, we propose the Uncertainty Guided Adaptive Prototype Network (UGAPNet) for semi-supervised few-shot semantic segmentation. The key innovation lies in the generation of reliable pseudo-prototypes as an additional supplement to alleviate intra-class semantic bias. Specifically, we employ a shared meta-learner to produce segmentation results for unlabeled images in the pseudo-label prediction module. Subsequently, we incorporate an uncertainty estimation module to quantify the difference between prototypes extracted from query and support images, facilitating pseudo-label denoising. Utilizing these refined pseudo-label samples, we introduce a prototype rectification module to obtain effective pseudo-prototypes and generate a generalized adaptive prototype for the segmentation of query images. Furthermore, generalized few-shot semantic segmentation extends the paradigm of few-shot semantic segmentation by simultaneously segmenting both unseen and seen classes during evaluation. To address the challenge of confusion region prediction between these two categories, we further propose a novel Prototype-Level Fusion Strategy in the prototypical contrastive space. Extensive experiments conducted on two benchmarks demonstrate the effectiveness of the proposed UGAPNet and prototype-level fusion strategy. Our source code will be available on https://github.com/WHL182/UGAPNet.

Evaluation of the reliability of measuring lower back muscles cross-sectional area based on manual segmentation within multi-level MRI images

IntroductionManual segmentation of paraspinal muscle cross-sectional area (CSA) is widely used to assess related health disorders. This study aimed to evaluate the reliability of this segmentation process for each paraspinal region of interest across the three intervertebral levels commonly used for segmentation (L3/4, L4/5, and L5/S1). MethodsAxial-T2-weighted MRI images for 238 patients were divided among five raters (47 ± 1 cases each). To conduct the intra-rater reliability study, the CSA of each paraspinal lumber muscle (psoas major (PM), multifidus (MF), and erector spinae (ES)) and the intervertebral disc (ID) were manually segmented twice on all targeted levels before being assessed for each region per rater. The Inter-rater reliability was determined by comparing the results of different readers who segmented the same dataset. The Intraclass Correlation (ICC) and Coefficient-of-Variation percentage (CV%) were reported for each analysis. ResultsLow intra- and inter-rater variability (CV%<11) was found for each reader in each region and intervertebral level. The inter-rater reliability was excellent (ICC>0.9) for the PM, ID, and ES at L3/4. However, it was very good for MF at all levels and ES at L4/5, L5/S1 (ICC range: 0.82–0.88) affected by the fat-infiltration nature of the ES and MF muscles, their proximity to each other, and their smaller size (correlation between muscle size and ICC = 0.6, P < 0.01). The ID segmentation has the lowest CV (<3 %) and excellent ICC (>0.93). ConclusionManual paraspinal muscle segmentation using axial-T2-weighted MRI is reliable at all commonly segmented intervertebral levels. However, the reliability level can be degraded by the presence of high-fat infiltrate, unclear muscle boundaries, and muscle size. Implications for practiceFollowing consistent guidelines can help improve segmentation results. The IDs can be used as reliable internal references.

Abnormal recognition-assisted and onset-offset aware network for pathological wearable ECG delineation

Electrocardiogram (ECG) delineation is essential to the identification of abnormal cardiac status, especially when ECG signals are remotely monitored with wearable devices. The complexity and diversity of cardiac conditions generate numerous pathological ECG patterns, not only requiring the recognition of normal ECG but also addressing an extensive range of abnormal ECG patterns, posing a challenging task. Therefore, we propose an abnormal recognition-assisted network to integrate supplementary information on diverse ECG patterns. Simultaneously, we design an onset-offset aware loss to enhance precise waveform localization. Specifically, we establish a two-branch framework where ECG delineation serves as the target task, producing the final segmentation results. Additionally, the abnormal recognition-assisted network serves as an auxiliary task, extracting multi-label pathological information from ECGs. This joint learning approach establishes crucial correlations between ECG delineation and associated ECG abnormalities. The correlations enable the model to demonstrate sufficient generalization in the presence of diverse abnormal ECG patterns. Besides, onset-offset aware loss focuses intensively on wave onsets and offsets by applying biased weights to various waveform positions. This approach ensures a focus on precise localization, facilitating seamless integration into cross-entropy loss function. A large-scale wearable 12-lead dataset containing 4,913 signals is collected, offering an extensive range of ECG data for model training. Results demonstrate that our method achieves outstanding performance on two test datasets, attaining sensitivity of 94.97% and 94.27% and an error tolerance lower than 20 ms. Furthermore, our method is effective for various aberrant ECG signals, including ST-segment changes, atrial premature beats, and right and left bundle branch blocks.

A progressive segmentation network for navigable areas with semantic–spatial information flow

Segmentation of safe navigable areas is a crucial technology for scene parsing in autopilot systems. However, existing segmentation methods often fail to adequately exploit the complementary relationships between multiscale features in complex wild environments, resulting in insufficient information fusion. In pursuit of a resolution, a Progressive Segmentation Network (PSNet) is proposed for navigable areas segmentation, which builds a semantic–spatial information flow branch to dynamically exploit the complementary relationships between multiscale features for progressive guided learning. Specifically, PSNet consists of four essential modules, including the Local Capturer and Global dependence Builder (LCGB), Multi-Directional Pooling Module (MDPM), Fusion-wise Module (FWM), and Spatial Weight Aggregation Module (SWAM). To facilitate efficient information dissemination, the LCGB was utilized to capture dense spatial information, and the MDPM was proposed to extract global geometric information of obstacles. These pieces of information serve as prior knowledge to guide learning. Additionally, we propose the FWM based on attention fusion unit (AFU) and contribution weights unit (CWU) to construct the complementary relationships between multiscale features and obtain rich multiscale fusion information. SWAM is proposed to enhance the significant spatial features from FWM and achieve impressive segmentation results. Extensive experimental results on the wild datasets demonstrate that PSNet outperforms state-of-the-art methods in recognizing safe navigable areas. The code for PSNet will be available at https://github.com/lv881314/PSNet.

Enhanced gastric cancer classification and quantification interpretable framework using digital histopathology images

Recent developments have highlighted the critical role that computer-aided diagnosis (CAD) systems play in analyzing whole-slide digital histopathology images for detecting gastric cancer (GC). We present a novel framework for gastric histology classification and segmentation (GHCS) that offers modest yet meaningful improvements over existing CAD models for GC classification and segmentation. Our methodology achieves marginal improvements over conventional deep learning (DL) and machine learning (ML) models by adaptively focusing on pertinent characteristics of images. This contributes significantly to our study, highlighting that the proposed model, which performs well on normalized images, is robust in certain respects, particularly in handling variability and generalizing to different datasets. We anticipate that this robustness will lead to better results across various datasets. An expectation-maximizing Naïve Bayes classifier that uses an updated Gaussian Mixture Model is at the heart of the suggested GHCS framework. The effectiveness of our classifier is demonstrated by experimental validation on two publicly available datasets, which produced exceptional classification accuracies of 98.87% and 97.28% on validation sets and 98.47% and 97.31% on test sets. Our framework shows a slight but consistent improvement over previously existing techniques in gastric histopathology image classification tasks, as demonstrated by comparative analysis. This may be attributed to its ability to capture critical features of gastric histopathology images better. Furthermore, using an improved Fuzzy c-means method, our study produces good results in GC histopathology picture segmentation, outperforming state-of-the-art segmentation models with a Dice coefficient of 65.21% and a Jaccard index of 60.24%. The model’s interpretability is complemented by Grad-CAM visualizations, which help understand the decision-making process and increase the model’s trustworthiness for end-users, especially clinicians.

MLFEU-NET: A Multi-scale Low-level Feature Enhancement Unet for breast lesions segmentation in ultrasound images

Breast lesions constantly threaten the health of females. Segmentation methods of breast lesions are important for clinical diagnosis, and neural networks have been widely used and played a significant role in this field. However, false detections and miss detections still exist due to the variability in the shape and location of breast lesions, artifacts and noise in breast ultrasound (BUS) images, and structural defects in conventional segmentation networks. In this paper, a multi-scale low-level feature enhancement Unet (MLFEU-net) structure is presented, consisting of the U-net structure, low-level feature enhancement block (LFEB), and a parallel multiscale feature fusion (PMFF) module. Specifically, LFEB enhances the detail information during shallow downsampling and further feature selection. Meanwhile, in the neck of Unet, PMFF module uses different scales of dilation convolution to provide different sizes of sensory fields, and to efficiently merge shallow and deep features filtered, leads to more accurate segmentation results. To evaluate the MLFEU-net’s segmentation ability, five quantitative metrics were used on two breast ultrasound datasets, and it was compared with several advanced segmentation techniques. The results illustrate that our approach surpasses other methods in performance. Moreover, robustness experiments validate the effectiveness of our approach in achieving robust segmentation of breast lesions.

A novel stone fruit quality non-destructive evaluation method based on PVR by LF-NMRI

For stone fruits, the pulp ratio, indicating the proportion of pulp to the total fruit, serves as a vital metric in assessing stone fruit quality. Traditionally, this ratio is computed by measuring the mass of each component through destructive sampling, and acquired results remain controversial. In order to obtain this ratio non-destructively and more accurately, a novel evaluation method for stone fruit quality based on pulp volume ratio (PVR) by low-field nuclear magnetic resonance imaging (LF-NMRI) is proposed. This approach integrates the SwinUnet segmentation network to differentiating the fruit pulp and core structure, thereby precisely computing the volume proportion of pulp. Experimental results reveal that SwinUnet exhibits better accuracy and robustness in the LF-NMRI data compared to the other networks. Based on the segmentation results, the disparities between the PVR values computed using the ellipse fitting method and the integral method, compared to the ground truth (GT), are less than 1% and 0.5% respectively. This paper provides a new reference for selecting premium stone fruits and enriches the fruit quality evaluation system.

A U-Net Model for Urban Land Cover Classification Using VHR Satellite Images

Urban settings are dynamic, constantly changing, and presenting a wide range of surface materials with high diversity in both spatial and spectral variation. As a result, mapping urban growth, evaluating infrastructure, managing water resources, and monitoring natural land cover become more complex tasks. Urban applications have made considerable progress thanks to the abundance of VHR orbital data and the recent development of artificial intelligence strategies especially neural networks. Convolutional neural networks have the potential to significantly enhance the analysis of urban land cover by addressing the limitations of traditional techniques. U-Net is a popular neural network for land cover analysis in remote sensing images. The current research presents a CNN model employing U-Net for image semantic segmentation in urban study area using both spectral and spatial context of VHR satellite data. The proposed model is trained, validated, and tested for VHR satellite image classification into five urban classes: water, vegetation, bare soil, road, and building. The CNN semantic segmentation results are compared to maximum likelihood image classification outcomes for validation and stability evaluation. A confusion matrix is applied to the classified scenes to determine the overall accuracy, producer's and user's accuracy, and Kappa coefficient using 400 random points with their corresponding ground truth. The U-Net image semantic segmentation technique achieved an overall accuracy of 87.50% and Kappa coefficient of 0.8395 which outperforms the maximum likelihood classification method with an overall accuracy of 83.25% and Kappa coefficient of 0.7812.

CMN: a co-designed neural architecture search for efficient computing-in-memory-based mixture-of-experts

Artificial intelligence (AI) has experienced substantial advancements recently, notably with the advent of large-scale language models (LLMs) employing mixture-of-experts (MoE) techniques, exhibiting human-like cognitive skills. As a promising hardware solution for edge MoE implementations, the computing-in-memory (CIM) architecture collocates memory and computing within a single device, significantly reducing the data movement and the associated energy consumption. However, due to diverse edge application scenarios and constraints, determining the optimal network structures for MoE, such as the expert’s location, quantity, and dimension on CIM systems remains elusive. To this end, we introduce a software-hardware co-designed neural architecture search (NAS) framework, CIM-based MoE NAS (CMN), focusing on identifying a high-performing MoE structure under specific hardware constraints. The results of the NYUD-v2 dataset segmentation on the RRAM (SRAM) CIM system reveal that CMN can discover optimized MoE configurations under energy, latency, and performance constraints, achieving 29.67× (43.10×) energy savings, 175.44×(109.89×) speedup, and 12.24× smaller model size compared to the baseline MoE-enabled Visual Transformer, respectively. This co-design opens up an avenue toward high-performance MoE deployments in edge CIM systems.

A multilevel lightweight fish respiratory frequency measurement method – Segmentation instead of detection

Respiratory frequency is an important physiological parameter for fish health. Traditional respiration detection methods have inherent frequency measurement error problems, and segmentation provides an innovative technique for respiration measurement. By shifting the focus from detection to segmentation, this approach provides robust and reliable support for measuring fish respiratory frequency accurately. However, the existing segmentation network models face challenges related to their large scale, high computational demands, and limited deployability at the edge. To overcome these challenges, this paper introduces an easily portable, high-precision segmentation method based on lightweight network and network slimming technique, named PM-YOLOv5s-seg (Prune MobileNetv3-YOLOv5s Segmentation). PM-YOLOv5s-seg utilizes the MobileNetv3 network as its segmentation backbone, enabling the extraction of essential fish mouth features. Simultaneously, it employs depth separable convolution to reduce network parameters and calculation amounts. Furthermore, the model undergoes additional sparsification via channel pruning, significantly improving its efficiency. Finally, respiratory frequency modeling is performed on the obtained fish mouth segmentation results, utilizing innovative techniques such as Ellipse fitting, Kalman filtering, and polynomial fitting methods. Comparison with the state-of-the-art methods demonstrates the superiority of the multilevel lightweight segmentation model PM-YOLOv5s-seg in terms of detection accuracy, computational efficiency, and model size. This model achieves an optimal balance between the speed and accuracy in fish respiratory frequency measurement, which holds significant implications for fish welfare as well as disease detection.

Ameliorated Fick’s law algorithm based multi-threshold medical image segmentation

Medical image segmentation is a critical and demanding step in medical image processing, which provides a solid foundation for subsequent medical image data extraction and analysis. Multi-threshold image segmentation, one of the most commonly used and specialized image segmentation techniques, limits its application to medical images because it requires demanding computational performance and is difficult to produce satisfactory segmentation results. To overcome the above problems, an ameliorated Fick's law algorithm (MsFLA) for multi-threshold image segmentation is developed in this paper. First, an optimized sine–cosine strategy is introduced to extend the molecular diffusion process to alleviate the problem of easily falling into local optima, thus improving the convergence accuracy of the Fick's law algorithm (FLA). Secondly, the introduction of local minimal value avoidance enriches the individual molecular information and enhances the local search ability, thus improving computational accuracy. In addition, the optimal neighborhood learning strategy is added to ensure a more careful and reasonable reliance on the optimal solution, thus reducing the chance of convergence of a local solution. The efficient optimization capability of MsFLA is comprehensively validated by comparing MsFLA with the original FLA and other algorithms in 23 classical benchmark functions. Finally, MsFLA is applied to image segmentation of grayscale images of COVID-19 and brain and color images of Lung and Colon cancer histopathology by using Cross entropy to validate its segmentation capability. The experimental results show that the MsFLA obtains the best segmentation results in three medical image cases compared to other comparison algorithms, which indicates that MsFLA can effectively solve the multi-threshold medical image segmentation problem.Graphical abstract

The Diachronic Annotated Corpus of Newar

Abstract This article presents the first Diachronic Annotated Corpus of Newar (DACON), an ongoing project to annotate Newar texts ranging from the 12th to the 19th centuries. The primary purpose of this corpus is to facilitate both synchronic and diachronic as well as comparative linguistic research. We also present our annotation pipeline with detailed information on our workflow and the results of the automatic segmentation and POS-tagging tools.

Federated learning via multi-attention guided UNet for thyroid nodule segmentation of ultrasound images

Accurate segmentation of thyroid nodules is essential for early screening and diagnosis, but it can be challenging due to the nodules' varying sizes and positions. To address this issue, we propose a multi-attention guided UNet (MAUNet) for thyroid nodule segmentation. We use a multi-scale cross attention (MSCA) module for initial image feature extraction. By integrating interactions between features at different scales, the impact of thyroid nodule shape and size on the segmentation results has been reduced. Additionally, we incorporate a dual attention (DA) module into the skip-connection step of the UNet network, which promotes information exchange and fusion between the encoder and decoder. To test the model's robustness and effectiveness, we conduct the extensive experiments on multi-center ultrasound images provided by 17 local hospitals. The model is trained using the federal learning mechanism to ensure privacy protection. The experimental results show that the Dice scores of the model on the data sets from the three centers are 0.908, 0.912 and 0.887, respectively. Compared to existing methods, our method demonstrates higher generalization ability on multi-center datasets and achieves better segmentation results.