Deep Learning Segmentation Methods Research Articles

Improved Robustness for Deep Learning-based Segmentation of Multi-Center Myocardial Perfusion MRI Datasets Using Data Adaptive Uncertainty-guided Space-time Analysis

Fully automatic analysis of myocardial perfusion MRI datasets enables rapid and objective reporting of stress/rest studies in patients with suspected ischemic heart disease. Developing deep learning techniques that can analyze multi-center datasets despite limited training data and variations in software (pulse sequence) and hardware (scanner vendor) is an ongoing challenge. Datasets from 3 medical centers acquired at 3T (n = 150 subjects; 21,150 first-pass images) were included: an internal dataset (inD; n = 95) and two external datasets (exDs; n = 55) used for evaluating the robustness of the trained deep neural network (DNN) models against differences in pulse sequence (exD-1) and scanner vendor (exD-2). A subset of inD (n = 85) was used for training/validation of a pool of DNNs for segmentation, all using the same spatiotemporal U-Net architecture and hyperparameters but with different parameter initializations. We employed a space-time sliding-patch analysis approach that automatically yields a pixel-wise "uncertainty map" as a byproduct of the segmentation process. In our approach, dubbed Data Adaptive Uncertainty-Guided Space-time (DAUGS) analysis, a given test case is segmented by all members of the DNN pool and the resulting uncertainty maps are leveraged to automatically select the "best" one among the pool of solutions. For comparison, we also trained a DNN using the established approach with the same settings (hyperparameters, data augmentation, etc.). The proposed DAUGS analysis approach performed similarly to the established approach on the internal dataset (Dice score for the testing subset of inD: 0.896 ± 0.050 vs. 0.890 ± 0.049; p = n.s.) whereas it significantly outperformed on the external datasets (Dice for exD-1: 0.885 ± 0.040 vs. 0.849 ± 0.065, p < 0.005; Dice for exD-2: 0.811 ± 0.070 vs. 0.728 ± 0.149, p < 0.005). Moreover, the number of image series with "failed" segmentation (defined as having myocardial contours that include bloodpool or are noncontiguous in ≥1 segment) was significantly lower for the proposed vs. the established approach (4.3% vs. 17.1%, p < 0.0005). The proposed DAUGS analysis approach has the potential to improve the robustness of deep learning methods for segmentation of multi-center stress perfusion datasets with variations in the choice of pulse sequence, site location or scanner vendor.

Projected pooling loss for red nucleus segmentation with soft topology constraints.

Deep learning is the standard for medical image segmentation. However, it may encounter difficulties when the training set is small. Also, it may generate anatomically aberrant segmentations. Anatomical knowledge can be potentially useful as a constraint in deep learning segmentation methods. We propose a loss function based on projected pooling to introduce soft topological constraints. Our main application is the segmentation of the red nucleus from quantitative susceptibility mapping (QSM) which is of interest in parkinsonian syndromes. This new loss function introduces soft constraints on the topology by magnifying small parts of the structure to segment to avoid that they are discarded in the segmentation process. To that purpose, we use projection of the structure onto the three planes and then use a series of MaxPooling operations with increasing kernel sizes. These operations are performed both for the ground truth and the prediction and the difference is computed to obtain the loss function. As a result, it can reduce topological errors as well as defects in the structure boundary. The approach is easy to implement and computationally efficient. When applied to the segmentation of the red nucleus from QSM data, the approach led to a very high accuracy (Dice 89.9%) and no topological errors. Moreover, the proposed loss function improved the Dice accuracy over the baseline when the training set was small. We also studied three tasks from the medical segmentation decathlon challenge (MSD) (heart, spleen, and hippocampus). For the MSD tasks, the Dice accuracies were similar for both approaches but the topological errors were reduced. We propose an effective method to automatically segment the red nucleus which is based on a new loss for introducing topology constraints in deep learning segmentation.

Brain Blood Vessel Segmentation based on Region Growing and U-net Neural Network

Aiming at the problems of insufficient feature extraction in existing deep learning-based cerebrovascular segmentation methods and uneven segmentation results when the original imaging quality is poor, a cerebrovascular segmentation method that integrates region growing and an improved U-net neural network is proposed. Based on the deep learning segmentation results, automatic seed point selection for region growing is performed and fused to improve the segmentation accuracy at the pixel level. In this work, the integrated segmentation method improves the Dice coefficient from 0.805 to 0.878 and the Average Hausdorff Distance (AHD) from 3.949 to 0.903 compared to the deep learning segmentation method.

GS-DeepLabV3+: A mountain tea disease segmentation network based on improved shuffle attention and gated multidimensional feature extraction

The prevention and control of Camellia Oleifera diseases can produce significantly comprehensive economic benefits. And precisely identifying the types of diseases on oil tea leaves can enhance the effectiveness of pest and disease control. Currently, the most effective deep learning methods for segmentation are suffering from the challenges such as indistinct disease spots on the actual oil tea leaves that are difficult to differentiate from the background, the high similarities in characteristics between different diseases, and the small target area of disease regions that are easily ignored. In this paper, GS-DeepLabV3+ network is proposed, which is an improvement upon DeepLabV3+ and aims to effectively enhance disease identification and segmentation rates. Initially, a gated pyramid feature fusion structure is proposed, utilizing a special module to merge features of different dimensions to enhance differentiation of similar features. Secondly, an improved grouped attention fusion mechanism is introduced, which allocates channel and spatial weights to different feature groups, enhancing the extraction of detailed features. Lastly, we substituted the backbone network of DeepLabV3+ with the more lightweight MobileNetV2, which significantly reduces the number of parameters while more stably extracting multidimensional features. In comparison with the existing segmentation networks on a self-created oil tea leaf disease data set, GS-DeepLabV3+ achieved a mIoU of 87.77% and an average pixel accuracy of 94.55%. Compared to the base DeepLabV3+ network, there was a 4.14% improvement in mIoU, a 3.58% increase in mPA, and a 1.4% rise in precision. The improved attention mechanism contributed to a 1.98% increase in mIoU, and the gated multidimensional fusion mechanism resulted in a 1.03% improvement. These enhancements can provide technical references for crop protection and pest disease segmentation.

Automated Scoring of Alzheimer's Disease Atrophy Scale with Subtype Classification Using Deep Learning-Based T1-Weighted Magnetic Resonance Image Segmentation.

Application of visual scoring scales for regional atrophy in Alzheimer's disease (AD) in clinical settings is limited by their high time cost and low intra/inter-rater agreement. To provide automated atrophy scoring using objective volume driven from deep-learning segmentation methods for AD subtype classification using magnetic resonance imaging (MRI). We enrolled 3,959 participants (1,732 cognitively normal [CN], 1594 with mild cognitive impairment [MCI], and 633 with AD). The occupancy indices for each regional volume were calculated by dividing each volume by the size of the lateral and inferior ventricular volumes. MR images from 355 participants (119 CN, 119 MCI, and 117 AD) from three different centers were used for validation. Two neuroradiologists performed visual assessments of the medial temporal, posterior, and global cortical atrophy scores in the frontal lobe using T1-weighted MR images. Images were also analyzed using the deep learning-based segmentation software, Neurophet AQUA. Cutoff values for the three scores were determined using the data distribution according to age. The scoring results were compared for consistency and reliability. Four volumetric-driven scoring results showed a high correlation with the visual scoring results for AD, MCI, and CN. The overall agreement with human raters was weak-to-moderate for atrophy scoring in CN participants, and good-to-almost perfect in AD and MCI participants. AD subtyping by automated scores also showed usefulness as a research tool. Determining AD subtypes using automated atrophy scoring for late-MCI and AD could be useful in clinical settings or multicenter studies with large datasets.

Innovative Point Cloud Segmentation of 3D Light Steel Framing System through Synthetic BIM and Mixed Reality Data: Advancing Construction Monitoring

In recent years, mixed reality (MR) technology has gained popularity in construction management due to its real-time visualisation capability to facilitate on-site decision-making tasks. The semantic segmentation of building components provides an attractive solution towards digital construction monitoring, reducing workloads through automation techniques. Nevertheless, data shortages remain an issue in maximizing the performance potential of deep learning segmentation methods. The primary aim of this study is to address this issue through synthetic data generation using Building Information Modelling (BIM) models. This study presents a point-cloud-based deep learning segmentation approach to a 3D light steel framing (LSF) system through synthetic BIM models and as-built data captured using MR headsets. A standardisation workflow between BIM and MR models was introduced to enable seamless data exchange across both domains. A total of five different experiments were set up to identify the benefits of synthetic BIM data in supplementing actual as-built data for model training. The results showed that the average testing accuracy using solely as-built data stood at 82.88%. Meanwhile, the introduction of synthetic BIM data into the training dataset led to an improved testing accuracy of 86.15%. A hybrid dataset also enabled the model to segment both the BIM and as-built data captured using an MR headset at an average accuracy of 79.55%. These findings indicate that synthetic BIM data have the potential to supplement actual data, reducing the costs associated with data acquisition. In addition, this study demonstrates that deep learning has the potential to automate construction monitoring tasks, aiding in the digitization of the construction industry.

Liver tumor segmentation method based on U-Net architecture: a review

Liver cancer is a disease with a high incidence and high probability of deterioration, and for the rapid diagnosis of liver disease, CT scans must be used to segment the liver tumors. For the past few years, with the rapid development of deep learning, many deep learning methods for liver tumor segmentation using abdominal computed tomography (CT) images have appeared, and the clinical application of these methods is of important significance for computer-aided diagnosis of liver tumors. The U-Net, with its unique U-shape network structure, exhibits excellent performance in medical image segmentation field and has been extensively utilized in various medical image segmentation applications. In this paper, we summarize the researches of U-Net and its improved networks in CT image segmentation of liver tumors by deep learning methods and classify various U-Net-based convolutional neural networks (CNNs) into 2D (two-dimensional), 3D (three-dimensional), and 2.5D (2.5-dimensional). In this paper, 2D, 3D, and 2.5D convolutional neural networks are summarized. In addition, this paper summarizes the advantages and disadvantages as well as the improvement methods of each type of network, which provides a useful reference for the studies of deep learning based on liver tumor segmentation field. Finally, this paper envisions future research trends for deep learning segmentation methods in the context of liver tumors.

Dense Multi-Scale Graph Convolutional Network for Knee Joint Cartilage Segmentation.

In this paper, we propose a dense multi-scale adaptive graph convolutional network (DMA-GCN) method for automatic segmentation of the knee joint cartilage from MR images. Under the multi-atlas setting, the suggested approach exhibits several novelties, as described in the following. First, our models integrate both local-level and global-level learning simultaneously. The local learning task aggregates spatial contextual information from aligned spatial neighborhoods of nodes, at multiple scales, while global learning explores pairwise affinities between nodes, located globally at different positions in the image. We propose two different structures of building models, whereby the local and global convolutional units are combined by following an alternating or a sequential manner. Secondly, based on the previous models, we develop the DMA-GCN network, by utilizing a densely connected architecture with residual skip connections. This is a deeper GCN structure, expanded over different block layers, thus being capable of providing more expressive node feature representations. Third, all units pertaining to the overall network are equipped with their individual adaptive graph learning mechanism, which allows the graph structures to be automatically learned during training. The proposed cartilage segmentation method is evaluated on the entire publicly available Osteoarthritis Initiative (OAI) cohort. To this end, we have devised a thorough experimental setup, with the goal of investigating the effect of several factors of our approach on the classification rates. Furthermore, we present exhaustive comparative results, considering traditional existing methods, six deep learning segmentation methods, and seven graph-based convolution methods, including the currently most representative models from this field. The obtained results demonstrate that the DMA-GCN outperforms all competing methods across all evaluation measures, providing DSC=95.71% and DSC=94.02% for the segmentation of femoral and tibial cartilage, respectively.

ESKNet: An enhanced adaptive selection kernel convolution for ultrasound breast tumors segmentation

Breast cancer has become one of the most dreaded diseases that can threaten the life of any woman. Accurate target lesion segmentation is essential for early clinical intervention and postoperative follow-up. Recently, many convolutional neural networks (CNNs) for segmenting breast tumors from ultrasound images have been presented. However, the complex ultrasound pattern and the variable tumor shape and size bring challenges to the accurate segmentation of the breast lesion. Motivated by the selective kernel convolution, we introduce an enhanced selective kernel convolution for breast tumor segmentation, which integrates multiple feature map region representations and adaptively recalibrates the weights of these feature map regions from the channel and spatial dimensions. This region recalibration strategy enables the network to focus more on high-contributing region features and mitigate the perturbation of less useful regions. Finally, the enhanced selective kernel convolution is integrated into U-net with deep supervision constraints to adaptively capture the robust representation of breast tumors. Using three public breast ultrasound datasets, we conducted extensive experiments with many state-of-the-art deep learning segmentation methods. In the segmentation of the first ultrasound dataset (BUSI), the values of Jaccard, Precision, Recall, Specificity and Dice are 70.20%, 79.57%, 82.41%, 97.47% and 78.71%, respectively. The values of Jaccard, Precision, Recall, Specificity and Dice for our method on the second ultrasound dataset (Dataset B) are 71.65%, 81.01%, 82.66%, 99.01% and 79.92%, respectively. For the segmentation of external ultrasound dataset (STU), the mean values of Jaccard, Precision, Recall, Specificity and Dice are 75.14%, 84.73%, 89.25%, 97.53% and 84.76%, respectively. The experimental results fully demonstrate the superior performance of our method for segmenting breast ultrasound images. The source code is available on the following website: https://github.com/CGPxy/ESKNet.

Automatic Quantitative Assessment for Diagnostic and Therapeutic Response in Rodent Myocardial Infarct Model.

The purpose of this study was to investigate the most appropriate methodological approach for the automatic measurement of rodent myocardial infarct polar map using histogram-based thresholding and unsupervised deep learning (DL)-based segmentation. A rat myocardial infarction model was induced by ligation of the left coronary artery. Positron emission tomography (PET) was performed 60 min after the administration of 18F-fluoro-deoxy-glucose (18F-FDG), and PET was performed after injecting 64Cu-pyruvaldehyde-bis(N4-methylthiosemicarbazone). Single photon emission computed tomography was performed 60 min after injection of 99mTc-hexakis-2-methoxyisobutylisonitrile and 201Tl. Delayed contrast-enhanced magnetic resonance imaging was performed after injecting Gd-DTPA-BMA. Three types of thresholding methods (naive thresholding, Otsu's algorithm, and multi-Gaussian mixture model (MGMM)) were used. DL segmentation methods were based on a convolution neural network and trained with constraints on feature similarity and spatial continuity of the response map extracted from images by the network. The relative infarct sizes measured by histology and estimated R2 for 18F-FDG were 0.8477, 0.7084, 0.8353, and 0.9024 for naïve thresholding, Otsu's algorithm, MGMM, and DL segmentation, respectively. DL-based method improved the accuracy of MI size assessment.

Impact of deep learning-based multiorgan segmentation methods on patient-specific internal dosimetry in PET/CT imaging: A comparative study.

Accurate and fast multiorgan segmentation is essential in image-based internal dosimetry in nuclear medicine. While conventional manual PET image segmentation is widely used, it suffers from both being time-consuming as well as subject to human error. This study exploited 2D and 3D deep learning (DL) models. Key organs in the trunk of the body were segmented and then used as a reference for networks. The pre-trained p2p-U-Net-GAN and HighRes3D architectures were fine-tuned with PET-only images as inputs. Additionally, the HighRes3D model was alternatively trained with PET/CT images. Evaluation metrics such as sensitivity (SEN), specificity (SPC), intersection over union (IoU), and Dice scores were considered to assess the performance of the networks. The impact of DL-assisted PET image segmentation methods was further assessed using the Monte Carlo (MC)-derived S-values to be used for internal dosimetry. A fair comparison with manual low-dose CT-aided segmentation of the PET images was also conducted. Although both 2D and 3D models performed well, the HighRes3D offers superior performance with Dice scores higher than 0.90. Key evaluation metrics such as SEN, SPC, and IoU vary between 0.89-0.93, 0.98-0.99, and 0.87-0.89 intervals, respectively, indicating the encouraging performance of the models. The percentage differences between the manual and DL segmentation methods in the calculated S-values varied between 0.1% and 6% with a maximum attributed to the stomach. The findings prove while the incorporation of anatomical information provided by the CT data offers superior performance in terms of Dice score, the performance of HighRes3D remains comparable without the extra CT channel. It is concluded that both proposed DL-based methods provide automated and fast segmentation of whole-body PET/CT images with promising evaluation metrics. Between them, the HighRes3D is more pronounced by providing better performance and can therefore be the method of choice for 18F-FDG-PET image segmentation.

PCNet: Prior Category Network for CT Universal Segmentation Model.

Accurate segmentation of anatomical structures in Computed Tomography (CT) images is crucial for clinical diagnosis, treatment planning, and disease monitoring. The present deep learning segmentation methods are hindered by factors such as data scale and model size. Inspired by how doctors identify tissues, we propose a novel approach, the Prior Category Network (PCNet), that boosts segmentation performance by leveraging prior knowledge between different categories of anatomical structures. Our PCNet comprises three key components: prior category prompt (PCP), hierarchy category system (HCS), and hierarchy category loss (HCL). PCP utilizes Contrastive Language-Image Pretraining (CLIP), along with attention modules, to systematically define the relationships between anatomical categories as identified by clinicians. HCS guides the segmentation model in distinguishing between specific organs, anatomical structures, and functional systems through hierarchical relationships. HCL serves as a consistency constraint, fortifying the directional guidance provided by HCS to enhance the segmentation model's accuracy and robustness. We conducted extensive experiments to validate the effectiveness of our approach, and the results indicate that PCNet can generate a high-performance, universal model for CT segmentation. The PCNet framework also demonstrates a significant transferability on multiple downstream tasks. The ablation experiments show that the methodology employed in constructing the HCS is of critical importance. The prompt and HCS can be accessed at https://github.com/YixinChen-AI/PCNet.

An overview of lab-based micro computed tomography aided finite element modelling of wood and its current bottlenecks

Abstract Microscopic lab-based X-ray computed tomography (XµCT) aided finite element (FE) modelling is a popular method with increasing nature within material science to predict local material properties of heterogeneous materials, e.g. elastic, hygroexpansion and diffusion. This method is relatively new to wood and lacks a clear methodology. Research intended to optimise the XµCT aided FE process often focuses on specific aspects within this process such as the XµCT scanning, segmentation or meshing, but not the entirety of the process. The compatibility and data transfer between aspects have not been investigated to the same extent, which creates errors that propagate and negatively impact the end results. In the current study, a methodology for the XµCT aided FE process of wood is suggested and its bottlenecks are identified based on a thorough literature review. Although the complexity of wood as a material makes it difficult to automate the XµCT aided FE process, the proposed methodology can assist in a more considered design and execution of this process. The main challenges that were identified include an automatic procedure to reconstruct the fibre orientation and to perform segmentation and meshing. A combined deep-learning segmentation method with geometry-based meshing can be suggested.

PCG-net: feature adaptive deep learning for automated head and neck organs-at-risk segmentation

IntroductionRadiation therapy is a common treatment option for Head and Neck Cancer (HNC), where the accurate segmentation of Head and Neck (HN) Organs-AtRisks (OARs) is critical for effective treatment planning. Manual labeling of HN OARs is time-consuming and subjective. Therefore, deep learning segmentation methods have been widely used. However, it is still a challenging task for HN OARs segmentation due to some small-sized OARs such as optic chiasm and optic nerve.MethodsTo address this challenge, we propose a parallel network architecture called PCG-Net, which incorporates both convolutional neural networks (CNN) and a Gate-Axial-Transformer (GAT) to effectively capture local information and global context. Additionally, we employ a cascade graph module (CGM) to enhance feature fusion through message-passing functions and information aggregation strategies. We conducted extensive experiments to evaluate the effectiveness of PCG-Net and its robustness in three different downstream tasks. ResultsThe results show that PCG-Net outperforms other methods, improves the accuracy of HN OARs segmentation, which can potentially improve treatment planning for HNC patients.DiscussionIn summary, the PCG-Net model effectively establishes the dependency between local information and global context and employs CGM to enhance feature fusion for accurate segment HN OARs. The results demonstrate the superiority of PCGNet over other methods, making it a promising approach for HNC treatment planning.

Deep-learning segmentation method for optical coherence tomography angiography in ophthalmology.

The optic disc and the macular are two major anatomical structures in the human eye. Optic discs are associated with the optic nerve. Macular mainly involves degeneration and impaired function of the macular region. Reliable optic disc and macular segmentation are necessary for the automated screening of retinal diseases. A swept-source OCTA system was designed to capture OCTA images of human eyes. To address these segmentation tasks, first, we constructed a new Optic Disc and Macula in fundus Image with optical coherence tomography angiography (OCTA) dataset (ODMI). Second, we proposed a Coarse and Fine Attention-Based Network (CFANet). The five metrics of our methods on ODMI are 98.91 , 98.47 , 89.77 , 98.49 , and 89.77 , respectively. Experimental results show that our CFANet has achieved good performance on segmentation for the optic disc and macula in OCTA.

Statin use and longitudinal changes in quantitative MRI-based biomarkers of thigh muscle quality: data from Osteoarthritis Initiative.

To assess whetherchanges in MRI-based measures of thigh muscle quality associated with statin use in participants with and without/at-risk of knee osteoarthritis. This retrospective cohort study used data from the Osteoarthritis Initiative study. Statin users and non-users were matched for relevant covariates using 1:1 propensity-score matching. Participants were further stratified according to baseline radiographic knee osteoarthritis status. We used a validated deep-learning method for thigh muscle MRI segmentation and calculation of muscle quality biomarkers at baseline, 2nd, and 4th visits. Mean difference and 95% confidence intervals (CI) in longitudinal 4-year measurements of muscle quality biomarkers, including cross-sectional area, intramuscular adipose tissue, contractile percent, and knee extensors and flexors maximum and specific contractile force (force/muscle area) were the outcomes of interest. After matching, 3772 thighs of 1910 participants were included (1886 thighs of statin-users: 1886 of non-users; age: 62 ± 9 years (average ± standard deviation), range: 45-79; female/male: 1). During 4 years, statin use was associated with a slight decrease in muscle quality, indicated by decreased knee extension maximum (mean-difference, 95% CI: - 1.85 N/year, - 3.23 to - 0.47) and specific contractile force (- 0.04 N/cm2/year, - 0.07 to - 0.01),decreased thigh muscle contractile percent (- 0.03%/year, - 0.06 to - 0.01), and increased thigh intramuscular adipose tissue (3.06 mm2/year, 0.53to 5.59). Stratified analyses showed decreased muscle quality only in participants without/at-risk of knee osteoarthritis but not those with established knee osteoarthritis. Statin use is associated with a slight decrease in MRI-based measures of thigh muscle quality over 4 years. However, considering statins' substantial cardiovascular benefits, these slight muscle changes may be relatively less important in overall patient care.

A deep learning segmentation method to assess dose to organs at risk during breast radiotherapy

Background and purposeContouring of organs at risk is important for studying health effects following breast radiotherapy. However, manual contouring is time-consuming and subject to variability. The purpose of this study was to develop a deep learning-based method to automatically segment multiple structures on breast radiotherapy planning computed tomography (CT) images. Materials and methodsWe used data from 118 patients, including 90 diagnostic CT scans with expert structure delineations for training and 28 breast radiotherapy planning CT images for testing. The radiotherapy CT images also had expert delineations for evaluating performance. We targeted a total of eleven organs at risk including five heart substructures. Segmentation performance was evaluated using the metrics of Dice similarity coefficient (DSC), overlap fraction, volume similarity, Hausdorff distance, mean surface distance, and dose. ResultsThe average DSC achieved on the radiotherapy planning images was 0.94 ± 0.02 for the whole heart, 0.96 ± 0.02 and 0.97 ± 0.01 for the left and right lung, 0.61 ± 0.10 for the esophagus, 0.81 ± 0.04 and 0.86 ± 0.04 for left and right atrium, 0.91 ± 0.02 and 0.84 ± 0.04 for left and right ventricle, and 0.21 ± 0.11 for the left anterior descending artery (LAD), respectively. Except for the LAD, the median difference in mean dose to these structures was small with absolute (relative) differences < 0.1 Gy (6 %). ConclusionsExcept for the LAD, our method demonstrated excellent performance and can be generalized to segment additional structures of interest.

BayeSeg: Bayesian modeling for medical image segmentation with interpretable generalizability.

Due to the cross-domain distribution shift aroused from diverse medical imaging systems, many deep learning segmentation methods fail to perform well on unseen data, which limits their real-world applicability. Recent works have shown the benefits of extracting domain-invariant representations on domain generalization. However, the interpretability of domain-invariant features remains a great challenge. To address this problem, we propose an interpretable Bayesian framework (BayeSeg) through Bayesian modeling of image and label statistics to enhance model generalizability for medical image segmentation. Specifically, we first decompose an image into a spatial-correlated variable and a spatial-variant variable, assigning hierarchical Bayesian priors to explicitly force them to model the domain-stable shape and domain-specific appearance information respectively. Then, we model the segmentation as a locally smooth variable only related to the shape. Finally, we develop a variational Bayesian framework to infer the posterior distributions of these explainable variables. The framework is implemented with neural networks, and thus is referred to as deep Bayesian segmentation. Quantitative and qualitative experimental results on prostate segmentation and cardiac segmentation tasks have shown the effectiveness of our proposed method. Moreover, we investigated the interpretability of BayeSeg by explaining the posteriors and analyzed certain factors that affect the generalization ability through further ablation studies. Our code is released via https://zmiclab.github.io/projects.html.

CartoCell, a high-content pipeline for 3D image analysis, unveils cell morphology patterns in epithelia

Decades of research have not yet fully explained the mechanisms of epithelial self-organization and 3D packing. Single-cell analysis of large 3D epithelial libraries is crucial for understanding the assembly and function of whole tissues. Combining 3D epithelial imaging with advanced deep-learning segmentation methods is essential for enabling this high-content analysis. We introduce CartoCell, a deep-learning-based pipeline that uses small datasets to generate accurate labels for hundreds of whole 3D epithelial cysts. Our method detects the realistic morphology of epithelial cells and their contacts in the 3D structure of the tissue. CartoCell enables the quantification of geometric and packing features at the cellular level. Our single-cell cartography approach then maps the distribution of these features on 2D plots and 3D surface maps, revealing cell morphology patterns in epithelial cysts. Additionally, we show that CartoCell can be adapted to other types of epithelial tissues.

Deep learning-based segmentation of whole-body fetal MRI and fetal weight estimation: assessing performance, repeatability, and reproducibility.

To develop a deep-learning method for whole-body fetal segmentation based on MRI; to assess the method's repeatability, reproducibility, and accuracy; to create an MRI-based normal fetal weight growth chart; and to assess the sensitivity to detect fetuses with growth restriction (FGR). Retrospective data of 348 fetuses with gestational age (GA) of 19-39weeks were included: 249 normal appropriate for GA (AGA), 19 FGR, and 80 Other (having various imaging abnormalities). A fetal whole-body segmentation model with a quality estimation module was developed and evaluated in 169 cases. The method was evaluated for its repeatability (repeated scans within the same scanner, n = 22), reproducibility (different scanners, n = 6), and accuracy (compared with birth weight, n = 7). A normal MRI-based growth chart was derived. The method achieved a Dice = 0.973, absolute volume difference ratio (VDR) = 1.8% and VDR mean difference = 0.75% ([Formula: see text]: - 3.95%, 5.46), and high agreement with the gold standard. The method achieved a repeatability coefficient = 4.01%, ICC = 0.99, high reproducibility with a mean difference = 2.21% ([Formula: see text]: - 1.92%, 6.35%), and high accuracy with a mean difference between estimated fetal weight (EFW) and birth weight of - 0.39% ([Formula: see text]: - 8.23%, 7.45%). A normal growth chart (n = 246) was consistent with four ultrasound charts. EFW based on MRI correctly predicted birth-weight percentiles for all 18 fetuses ≤ 10thpercentile and for 14 out of 17 FGR fetuses below the 3rd percentile. Six fetuses referred to MRI as AGA were found to be < 3rd percentile. The proposed method for automatic MRI-based EFW demonstrated high performance and sensitivity to identify FGR fetuses. Results from this study support the use of the automatic fetal weight estimation method based on MRI for the assessment of fetal development and to detect fetuses at risk for growth restriction. • An AI-based segmentation method with a quality assessment module for fetal weight estimation based on MRI was developed, achieving high repeatability, reproducibility, and accuracy. • An MRI-based fetal weight growth chart constructed from a large cohort of normal and appropriate gestational-age fetuses is proposed. • The method showed a high sensitivity for the diagnosis of small fetuses suspected of growth restriction.