Medical image analysis最新文献

Automatic 3-dimensional tooth segmentation on intraoral scans (IOS) plays a pivotal role in computer-aided orthodontic treatments. In practice, deploying existing well-trained models to different medical centers suffers from two main problems: (1) the data distribution shifts between existing and new centers, which causes significant performance degradation. (2) The data in the existing center(s) is usually not permitted to be shared, and annotating additional data in the new center(s) is time-consuming and expensive, thus making re-training or fine-tuning unfeasible. In this paper, we propose a framework for Cross-center Model Adaptive Tooth segmentation (CMAT) to alleviate these issues. CMAT takes the trained model(s) from the source center(s) as input and adapts them to different target centers, without data transmission or additional annotations. CMAT is applicable to three cross-center scenarios: source-data-free, multi-source-data-free, and test-time. The model adaptation in CMAT is realized by a tooth-level prototype alignment module, a progressive pseudo-labeling transfer module, and a tooth-prior regularized information maximization module. Experiments under three cross-center scenarios on two datasets show that CMAT can consistently surpass existing baselines. The effectiveness is further verified with extensive ablation studies and statistical analysis, demonstrating its applicability for privacy-preserving model adaptive tooth segmentation in real-world digital dentistry.

Recent years have seen a growing interest in methods for predicting an unknown variable of interest, such as a subject's diagnosis, from medical images depicting its anatomical-functional effects. Methods based on discriminative modeling excel at making accurate predictions, but are challenged in their ability to explain their decisions in anatomically meaningful terms. In this paper, we propose a simple technique for single-subject prediction that is inherently interpretable. It augments the generative models used in classical human brain mapping techniques, in which the underlying cause-effect relations can be encoded, with a multivariate noise model that captures dominant spatial correlations. Experiments demonstrate that the resulting model can be efficiently inverted to make accurate subject-level predictions, while at the same time offering intuitive visual explanations of its inner workings. The method is easy to use: training is fast for typical training set sizes, and only a single hyperparameter needs to be set by the user. Our code is available at https://github.com/chiara-mauri/Interpretable-subject-level-prediction.

Variations in hue and contrast are common in H&E-stained pathology images due to differences in slide preparation across various institutions. Such stain variations, while not affecting pathologists much in diagnosing the biopsy, pose significant challenges for computer-assisted diagnostic systems, leading to potential underdiagnosis or misdiagnosis, especially when stain differentiation introduces substantial heterogeneity across datasets from different sources. Traditional stain normalization methods, aimed at mitigating these issues, often require labor-intensive selection of appropriate templates, limiting their practicality and automation. Innovatively, we propose a Learnable Stain Normalization layer, i.e. LStainNorm, designed as an easily integrable component for pathology image analysis. It minimizes the need for manual template selection by autonomously learning the optimal stain characteristics. Moreover, the learned optimal stain template provides the interpretability to enhance the understanding of the normalization process. Additionally, we demonstrate that fusing pathology images normalized in multiple color spaces can improve performance. Therefore, we extend LStainNorm with a novel self-attention mechanism to facilitate the fusion of features across different attributes and color spaces. Experimentally, LStainNorm outperforms the state-of-the-art methods including conventional ones and GANs on two classification datasets and three nuclei segmentation datasets by an average increase of 4.78% in accuracy, 3.53% in Dice coefficient, and 6.59% in IoU. Additionally, by enabling an end-to-end training and inference process, LStainNorm eliminates the need for intermediate steps between normalization and analysis, resulting in more efficient use of hardware resources and significantly faster inference time, i.e up to hundreds of times quicker than traditional methods. The code is publicly available at https://github.com/yjzscode/Optimal-Normalisation-in-Color-Spaces.

Accurate tool tracking is essential for the success of computer-assisted intervention. Previous efforts often modeled tool trajectories rigidly, overlooking the dynamic nature of surgical procedures, especially tracking scenarios like out-of-body and out-of-camera views. Addressing this limitation, the new CholecTrack20 dataset provides detailed labels that account for multiple tool trajectories in three perspectives: (1) intraoperative, (2) intracorporeal, and (3) visibility, representing the different types of temporal duration of tool tracks. These fine-grained labels enhance tracking flexibility but also increase the task complexity. Re-identifying tools after occlusion or re-insertion into the body remains challenging due to high visual similarity, especially among tools of the same category. This work recognizes the critical role of the tool operators in distinguishing tool track instances, especially those belonging to the same tool category. The operators' information are however not explicitly captured in surgical videos. We therefore propose SurgiTrack, a novel deep learning method that leverages YOLOv7 for precise tool detection and employs an attention mechanism to model the originating direction of the tools, as a proxy to their operators, for tool re-identification. To handle diverse tool trajectory perspectives, SurgiTrack employs a harmonizing bipartite matching graph, minimizing conflicts and ensuring accurate tool identity association. Experimental results on CholecTrack20 demonstrate SurgiTrack's effectiveness, outperforming baselines and state-of-the-art methods with real-time inference capability. This work sets a new standard in surgical tool tracking, providing dynamic trajectories for more adaptable and precise assistance in minimally invasive surgeries.

Designing dental crowns with computer-aided design software in dental laboratories is complex and time-consuming. Using real clinical datasets, we developed an end-to-end deep learning model that automatically generates personalized dental crown meshes. The input context includes the prepared tooth, its adjacent teeth, and the two closest teeth in the opposing jaw. The training set contains this context, the ground truth crown, and the extracted margin line. Our model consists of two components: First, a feature extractor converts the input point cloud into a set of local feature vectors, which are then fed into a transformer-based model to predict the geometric features of the crown. Second, a point-to-mesh module generates a dense array of points with normal vectors, and a differentiable Poisson surface reconstruction method produces an accurate crown mesh. Training is conducted with three losses: (1) a customized margin line loss; (2) a contrastive-based Chamfer distance loss; and (3) a mean square error (MSE) loss to control mesh quality. We compare our method with our previously published method, Dental Mesh Completion (DMC). Extensive testing confirms our method's superiority, achieving a 12.32% reduction in Chamfer distance and a 46.43% reduction in MSE compared to DMC. Margin line loss improves Chamfer distance by 5.59%.

Resting-state functional magnetic resonance imaging (rsfMRI) is a powerful tool for investigating the relationship between brain function and cognitive processes as it allows for the functional organization of the brain to be captured without relying on a specific task or stimuli. In this paper, we present a novel modeling architecture called BrainRGIN for predicting intelligence (fluid, crystallized and total intelligence) using graph neural networks on rsfMRI derived static functional network connectivity matrices. Extending from the existing graph convolution networks, our approach incorporates a clustering-based embedding and graph isomorphism network in the graph convolutional layer to reflect the nature of the brain sub-network organization and efficient network expression, in combination with TopK pooling and attention-based readout functions. We evaluated our proposed architecture on a large dataset, specifically the Adolescent Brain Cognitive Development Dataset, and demonstrated its effectiveness in predicting individual differences in intelligence. Our model achieved lower mean squared errors and higher correlation scores than existing relevant graph architectures and other traditional machine learning models for all of the intelligence prediction tasks. The middle frontal gyrus exhibited a significant contribution to both fluid and crystallized intelligence, suggesting their pivotal role in these cognitive processes. Total composite scores identified a diverse set of brain regions to be relevant which underscores the complex nature of total intelligence. Our GitHub implementation is publicly available on https://github.com/bishalth01/BrainRGIN/.

Computed tomography (CT) is continuously becoming a valuable diagnostic technique in clinical practice. However, the radiation dose exposure in the CT scanning process is a public health concern. Within medical diagnoses, mitigating the radiation risk to patients can be achieved by reducing the radiation dose through adjustments in tube current and/or the number of projections. Nevertheless, dose reduction introduces additional noise and artifacts, which have extremely detrimental effects on clinical diagnosis and subsequent analysis. In recent years, the feasibility of applying deep learning methods to low-dose CT (LDCT) imaging has been demonstrated, leading to significant achievements. This article proposes a dual-domain joint optimization LDCT imaging framework (termed DDoCT) which uses noisy sparse-view projection to reconstruct high-performance CT images with joint optimization in projection and image domains. The proposed method not only addresses the noise introduced by reducing tube current, but also pays special attention to issues such as streak artifacts caused by a reduction in the number of projections, enhancing the applicability of DDoCT in practical fast LDCT imaging environments. Experimental results have demonstrated that DDoCT has made significant progress in reducing noise and streak artifacts and enhancing the contrast and clarity of the images.

Coronary artery disease (CAD) is the leading cause of death globally. The 3D fusion of coronary X-ray angiography (XA) and optical coherence tomography (OCT) provides complementary information to appreciate coronary anatomy and plaque morphology. This significantly improve CAD diagnosis and prognosis by enabling precise hemodynamic and computational physiology assessments. The challenges of fusion lie in the potential misalignment caused by the foreshortening effect in XA and non-uniform acquisition of OCT pullback. Moreover, the need for reconstructions of major bifurcations is technically demanding. This paper proposed an automated 3D fusion framework AutoFOX, which consists of deep learning model TransCAN for 3D vessel alignment. The 3D vessel contours are processed as sequential data, whose features are extracted and integrated with bifurcation information to enhance alignment via a multi-task fashion. TransCAN shows the highest alignment accuracy among all methods with a mean alignment error of 0.99 ± 0.81 mm along the vascular sequence, and only 0.82 ± 0.69 mm at key anatomical positions. The proposed AutoFOX framework uniquely employs an advanced side branch lumen reconstruction algorithm to enhance the assessment of bifurcation lesions. A multi-center dataset is utilized for independent external validation, using the paired 3D coronary computer tomography angiography (CTA) as the reference standard. Novel morphological metrics are proposed to evaluate the fusion accuracy. Our experiments show that the fusion model generated by AutoFOX exhibits high morphological consistency with CTA. AutoFOX framework enables automatic and comprehensive assessment of CAD, especially for the accurate assessment of bifurcation stenosis, which is of clinical value to guiding procedure and optimization.

Photoacoustic tomography (PAT), as a novel biomedical imaging technique, is able to capture temporal, spatial and spectral tomographic information from organisms. Organ-level multi-parametric analysis of continuous PAT images are of interest since it enables the quantification of organ specific morphological and functional parameters in small animals. Accurate organ delineation is imperative for organ-level image analysis, yet the low contrast and blurred organ boundaries in PAT images pose challenge for their precise segmentation. Fortunately, shared structural information among continuous images in the time-space-spectrum domain may be used to enhance segmentation. In this paper, we introduce a structure fusion enhanced graph convolutional network (SFE-GCN), which aims at automatically segmenting major organs including the body, liver, kidneys, spleen, vessel and spine of abdominal PAT image of mice. SFE-GCN enhances the structural feature of organs by fusing information in continuous image sequence captured at time, space and spectrum domains. As validated on large-scale datasets across different imaging scenarios, our method not only preserves fine structural details but also ensures anatomically aligned organ contours. Most importantly, this study explores the application of SFE-GCN in multi-dimensional organ image analysis, including organ-based dynamic morphological analysis, organ-wise light fluence correction and segmentation-enhanced spectral un-mixing. Code will be released at https://github.com/lzc-smu/SFEGCN.git.