To address the limitations of transductive learning in evolving real-world scenarios where unknown categories may continuously emerge, Continual Generalized Category Discovery (C-GCD) presents a novel paradigm that extends conventional category discovery frameworks. Unlike traditional static learning environments, C-GCD requires models to incrementally discover novel categories across multiple operational phases while maintaining discrimination capabilities for previously learned classes, posing significant challenges in balancing stability and plasticity. Prior approaches typically employ parameter-level knowledge distillation from historical models to alleviate catastrophic forgetting, which effectively preserves prior knowledge and optimizes computational efficiency. However, our analysis reveals that the persistent availability of samples from previous stages enables more sophisticated knowledge preservation strategies. Specifically, we present a Fix and Explore strategy that employs distinct learning methodologies for different types of potential data, aiming to preserve the features of old categories as much as possible and gradually exploring the potential distribution of new class latent spaces, we can enhance the model's ability to discover novel categories. This paper investigates this effect and introduces a novel heuristic paradigm to solve the C-GCD problem, called Fix and Explore (FaE), which aims to provide sufficient imaginative space for new classes while preserving the classification ability for old tasks. We conducted experiments across multiple datasets and performed detailed comparisons. The results demonstrate that our method achieves state-of-the-art performance at each stage across all datasets.

Developing heart sound classification algorithms for murmur identification is critical for early screening of heart diseases. However, identifying murmurs in long-duration heart sound signals can be challenging due to their weak features and interference from noise. Considering the periodic patterns of heart sounds and murmurs, periodic priors can be introduced to enhance murmur identification, an approach that remains underutilized in current methods. In this study, we propose a novel Dual-level Periodic Pattern-Aware Transformer (DPPAT) to implicitly leverage the periodic priors of heart sound signals without requiring cycle segmentation. In the regional-level, an Adaptive Period-Aligned Window Selection algorithm is designed for the model to extract periodic components while suppressing random noise using a Periodic Pattern Attention module. In the global-level, the model further integrates these periodic features in global-modeling to enhance the identification of murmur-discriminative features. Validated on the dataset from 2022 George B. Moody PhysioNet Challenge, our proposed method achieves a weighted accuracy of 84.27% and an F1-score of 70.38% through 10-fold cross-validation. The generalizability of DPPAT is further verified on two additional public datasets, including both heart sound and respiratory sound signals. Furthermore, attention visualizations provide a clear understanding of the focus of the model, highlighting the decision-making basis for murmur identification.

Real-time guidewire tracking and position estimation are crucial for intraoperative navigation during percutaneous coronary intervention (PCI). Compared with other interventional procedures, PCI employs smaller-sized guidewires that are extremely difficult to distinguish from surrounding anatomical structures due to their slenderness and complex anatomical backgrounds. Moreover, guidewires typically exhibit low signal-to-noise ratios (SNR) in fluoroscopic video sequences. Additionally, complex motion artifacts caused by patient breathing and cardiac movements further complicate real-time guidewire localization. To address these challenges, we propose a novel end-to-end framework for guidewire segmentation and localization. Given that background information in PCI images predominantly occupies low-frequency components, we designed a multi-scale feature aggregation module based on low-frequency suppression to enhance the network's capability for extracting slender structures such as guidewires. Furthermore, to handle discontinuities caused by low SNR and structural occlusions in preliminary segmentation results, we incorporated a lightweight reconstruction network along with a generative adversarial network (GAN) to repair broken regions. The two networks are trained jointly, and their outputs are ultimately fused to achieve accurate guidewire segmentation. Extensive experiments conducted on multi-center private datasets demonstrate the superior performance of our approach, with F1 score of 76.45% and a localization precision of 0.18 mm. With a compact model size of only 12.64 M parameters and a real-time processing speed of 46 FPS, our method offers a highly promising solution for future clinical applications.

The recently proposed Novel Category Discovery (NCD) adapt paradigm of transductive learning hinders its application in more real-world scenarios. In fact, few labeled data in part of new categories can well alleviate this burden, which coincides with the ease that people can label few of new category data. Therefore, this paper presents a new setting in which a trained agent is able to flexibly switch between the tasks of identifying examples of known (labelled) classes and clustering novel (completely unlabeled) classes as the number of query examples increases by leveraging knowledge learned from only a few (handful) support examples. Drawing inspiration from the discovery of novel categories using prior-based clustering algorithms, we introduce a novel framework that further relaxes its assumptions to the real-world open set level by unifying the concept of model adaptability in few-shot learning. We refer to this setting as Few-Shot Novel Category Discovery (FSNCD) and propose Semi-supervised Hierarchical Clustering (SHC) and Uncertainty-aware K-means Clustering (UKC) to examine the model's reasoning capabilities. Extensive experiments and detailed analysis on five commonly used datasets demonstrate that our methods can achieve leading performance levels across different task settings and scenarios. Code is available at: https://github.com/Ashengl/FSNCD.

The registration of coronary artery structures from preoperative coronary computed tomography angiography to intraoperative coronary angiography is of great interest to improve guidance in percutaneous coronary interventions. However, non-rigid deformation and discrepancies in both dimensions and topology between the two imaging modalities present a challenge in the 2D/3D coronary artery registration. In this study, we address this problem by formulating it as a centerline feature matching task and propose a GNN-based vessel matching network (GVM-Net) to establish dense correspondence between different image modalities in an end-to-end manner. GVM-Net considers centerline points as nodes in graphs and effectively models the complex topological relationships between them through attention mechanisms and message passing. Furthermore, by incorporating redundant rows and columns into the matching matrix, GVM-Net can effectively handle inconsistencies in vascular structures. We also introduce the query-based nodes grouping module, which clusters nodes in the feature space to further explore the topological relationships. GVM-Net achieves an average F1-score of 89.74% with a mean pixel distance of 0.48 pixels on the synthetic dataset with 276 data pairs and an average F1-score of 83.35% with a mean error of 1.52 mm in 55 manually labeled clinical cases, both exceeding existing feature matching methods.

Unmanned equipment is more and more widely used in complex dangerous scenarios such as emergency rescue. The cooperative operation of manned/unmanned hybrid formation will be the main method at present. This paper proposes a topology analysis and evaluation method of manned/unmanned cooperative equipment systems and constructs a topology model from the logic layer to the physical layer. According to the concepts of degree and betweenness centrality, the key nodes analysis and invulnerability evaluation method of manned/unmanned equipment system are proposed. Finally, the method's effectiveness is verified by simulation, which provides system and data support for dynamic reconstruction and real-time decision-making of manned/unmanned equipment systems.

A regenerative braking control strategy is proposed for an electrical-four-wheel-drive (e-4WD) hybrid vehicle. The e-4WD powertrain consists of an engine, a motor/generator (MG), and an automatic transmission (AT) on the front axle, and two in-wheel motors (IWMs) on the rear axle. Pure regenerative braking with the MG and two IWMs is focused in this article, which is very common in mild braking scenario. Energy efficiency and vehicle motion stability are both considered in the design of the braking control strategy. With quadratic programming (QP)-based control allocation algorithm, downshifting decision of the AT is coordinated with the torque allocation of the MG and two IWMs, achieving high regenerative energy efficiency as well as good vehicle motion dynamics. Co-simulations with Matlab/Simulink®and CarSim®show the effectiveness of the proposed algorithm.

Geometric dynamic configurations of body joints play an essential role in distinguishing different human activities. However, many existing human activity recognition approaches lack the capability of automatically learning these configurations from sequences of joints in four-dimensional space (spatio and temporal). In this paper, the authors propose an automatic joint configuration learning method, based on dictionary learning and sparse representation. The proposed method achieves the following features: 1) it automatically learns dynamic spatio-temporal geometric configurations of body joints, involved in activities, in a simple way; 2) it dispenses with the hand crafted feature designing process and provides a new method to organize joint coordinate data as fixed length column vectors, which are suitable for dictionary learning; 3) it replaces the conventional bag of words model with sparse coding method; words in learned dictionary capture subactivity features, and the frequencies of different words appearing in different activities characterize the categories of global activity; 4) it is robust to time misalignment and can classify any length of video sequence (online classification) in real time; 5) it is easy to combine this method with other forms of data for better performance, because of its data driven nature and flexible framework. The proposed method is tested with three state-of-the-art public human activity recognition datasets and the results are found to be better than those of CAD-60 dataset, and comparable to those of both MSR Action 3D and MSR Daily Activity datasets (source codes are publicly available at https://github.com/jinqijinqi/SparseCodingDictionaryLearningHumanActivityRecognition).

In this paper, a level set method is proposed for the segmentation of Left Ventricle (LV) from short-axis cardiac magnetic resonance images. According to the anatomical knowledge of LV, we first propose a convexity preserving mechanism to keep the shape of the evolving contour convex during the curve evolution, and thereby improves the segmentation accuracy. Then, the mechanism is incorporated into two-layer level set method to delineate endocardial and epicardial boundaries simultaneously. The proposed method has been quantitatively validated on a public dataset, and experimental results and comparisons with other methods demonstrate the superior performance of our method. Furthermore, such a generally constrained convexity-preserving level set method can be useful in many other potential applications, as validated by experiments.

The detection of multiple sclerosis lesion is important for many neuroimaging studies. In this paper, a new automatic algorithm for lesion segmentation based on the multi-channel MR images (T1w, T2w and FLAIR image) is proposed, which utilizes the unique and complementary intensity information of multi-channel MR images. In this method, the observed multi-channel MR images are modeled as a vector valued image. The image in each channel consists of two multiplicative components: a smooth varying bias filed vector and a piecewise approximately constant true image vector. An energy function of this vector valued image is defined in term of the property of true image and bias field. The energy minimization is proposed for seeking the optimal segmentation result of lesions. Our method is applied to the real multi-channel MR images and compared with two sets of manual segmentation followed by the quantitative evaluation. The experimental results show that our approach is effective and robust for the lesion segmentation.

Intensity inhomogeneity often occurs in real-world images, which presents a considerable challenge in image segmentation. The most widely used image segmentation algorithms are region-based and typically rely on the homogeneity of the image intensities in the regions of interest, which often fail to provide accurate segmentation results due to the intensity inhomogeneity. This paper proposes a novel region-based method for image segmentation, which is able to deal with intensity inhomogeneities in the segmentation. First, based on the model of images with intensity inhomogeneities, we derive a local intensity clustering property of the image intensities, and define a local clustering criterion function for the image intensities in a neighborhood of each point. This local clustering criterion function is then integrated with respect to the neighborhood center to give a global criterion of image segmentation. In a level set formulation, this criterion defines an energy in terms of the level set functions that represent a partition of the image domain and a bias field that accounts for the intensity inhomogeneity of the image. Therefore, by minimizing this energy, our method is able to simultaneously segment the image and estimate the bias field, and the estimated bias field can be used for intensity inhomogeneity correction (or bias correction). Our method has been validated on synthetic images and real images of various modalities, with desirable performance in the presence of intensity inhomogeneities. Experiments show that our method is more robust to initialization, faster and more accurate than the well-known piecewise smooth model. As an application, our method has been used for segmentation and bias correction of magnetic resonance (MR) images with promising results.

Level set methods have been widely used in image processing and computer vision. In conventional level set formulations, the level set function typically develops irregularities during its evolution, which may cause numerical errors and eventually destroy the stability of the evolution. Therefore, a numerical remedy, called reinitialization, is typically applied to periodically replace the degraded level set function with a signed distance function. However, the practice of reinitialization not only raises serious problems as when and how it should be performed, but also affects numerical accuracy in an undesirable way. This paper proposes a new variational level set formulation in which the regularity of the level set function is intrinsically maintained during the level set evolution. The level set evolution is derived as the gradient flow that minimizes an energy functional with a distance regularization term and an external energy that drives the motion of the zero level set toward desired locations. The distance regularization term is defined with a potential function such that the derived level set evolution has a unique forward-and-backward (FAB) diffusion effect, which is able to maintain a desired shape of the level set function, particularly a signed distance profile near the zero level set. This yields a new type of level set evolution called distance regularized level set evolution (DRLSE). The distance regularization effect eliminates the need for reinitialization and thereby avoids its induced numerical errors. In contrast to complicated implementations of conventional level set formulations, a simpler and more efficient finite difference scheme can be used to implement the DRLSE formulation. DRLSE also allows the use of more general and efficient initialization of the level set function. In its numerical implementation, relatively large time steps can be used in the finite difference scheme to reduce the number of iterations, while ensuring sufficient numerical accuracy. To demonstrate the effectiveness of the DRLSE formulation, we apply it to an edge-based active contour model for image segmentation, and provide a simple narrowband implementation to greatly reduce computational cost.

This paper proposes a new energy minimization framework for simultaneous estimation of the bias field and segmentation of tissues for magnetic resonance images. The bias field is modeled as a linear combination of a set of basis functions, and thereby parameterized by the coefficients of the basis functions. We define an energy that depends on the coefficients of the basis functions, the membership functions of the tissues in the image, and the constants approximating the true signal from the corresponding tissues. This energy is convex in each of its variables. Bias field estimation and image segmentation are simultaneously achieved as the result of minimizing this energy. We provide an efficient iterative algorithm for energy minimization, which converges to the optimal solution at a fast rate. A salient advantage of our method is that its result is independent of initialization, which allows robust and fully automated application. The proposed method has been successfully applied to 3-Tesla MR images with desirable results. Comparisons with other approaches demonstrate the superior performance of this algorithm.

Spatial intensity variations caused by illumination changes have been a challenge for image segmentation and many other computer vision tasks. This paper presents a novel method for image segmentation with simultaneous estimation of illumination and reflectance images. The proposed method is based on the composition of an observed scene image with an illumination component and a reflectance component, known as intrinsic images. We define an energy functional in terms of an illumination image, the membership functions of the regions, and the corresponding reflectance constants of the regions in the scene. This energy is convex in each of its variables. By minimizing the energy, image segmentation result is obtained in the form of the membership functions of the regions. The illumination and reflectance components of the observed image are estimated simultaneously as the result of energy minimization. With illumination taken into account, the proposed method is able to segment images with non-uniform intensities caused by spatial variations in illumination. Comparisons with the state-of-the-art piecewise smooth model demonstrate the superior performance of our method.

This paper presents a new image denoising model for real color photo noise removal. Our model is implemented in the hue, saturation and intensity (HSI) space. The hue and saturation denoising are combined and implemented as a complex total variation (TV) diffusion. The intensity denoising is based on a diffusion flow to minimize a new energy functional, which is constructed with intensity component statistics. Besides the common gradient-based edge stopping functions for anisotropic diffusion, specifically for color photo denoising, we incorporate an intensity-based brightness adjusting term in the new energy, which corresponds to the noise disturbance with respect to photo intensity. In addition, we use the gradient vector flow (GVF) as the new diffusion directions for more accurate and robust denoising. Compared with previous diffusion flows only based on regular image gradients, this model provides more accurate image structure and intensity noise characterization for better denoising. Comprehensive quantitative and qualitative experiments on color photos demonstrate the improved performance of the proposed model when compared with 14 recognized approaches and 2 commercial software.

This paper presents a novel level set method for image segmentation. Gray-level moments are used to estimate two fitting functions that approximate local intensities on the two sides of object boundaries, which are then incorporated into a variational level set framework. An energy functional is defined on a contour, which characterizes the approximation of local intensities on the two sides of the contour by the two fitting functions. This energy can be minimized when the contour is on the object boundary. Thus, image segmentation is performed by minimizing this energy functional. A desirable feature of our method is that it is not sensitive to the contour initialization. Moreover, our method is able to segment images with intensity inhomogeneity. Only a small number of iterations are needed to obtain the final result, which makes our method more efficient than previous level set methods.

Intensity inhomogeneities often occur in real-world images and may cause considerable difficulties in image segmentation. In order to overcome the difficulties caused by intensity inhomogeneities, we propose a region-based active contour model that draws upon intensity information in local regions at a controllable scale. A data fitting energy is defined in terms of a contour and two fitting functions that locally approximate the image intensities on the two sides of the contour. This energy is then incorporated into a variational level set formulation with a level set regularization term, from which a curve evolution equation is derived for energy minimization. Due to a kernel function in the data fitting term, intensity information in local regions is extracted to guide the motion of the contour, which thereby enables our model to cope with intensity inhomogeneity. In addition, the regularity of the level set function is intrinsically preserved by the level set regularization term to ensure accurate computation and avoids expensive reinitialization of the evolving level set function. Experimental results for synthetic and real images show desirable performances of our method.

Local image information is crucial for accurate segmentation of images with intensity inhomogeneity. However, image information in local region is not embedded in popular region-based active contour models, such as the piecewise constant models. In this paper, we propose a region-based active contour model that is able to utilize image information in local regions. The major contribution of this paper is the introduction of a local binary fitting energy with a kernel function, which enables the extraction of accurate local image information. Therefore, our model can be used to segment images with intensity inhomogeneity, which overcomes the limitation of piecewise constant models. Comparisons with other major region-based models, such as the piece-wise smooth model, show the advantages of our method in terms of computational efficiency and accuracy. In addition, the proposed method has promising application to image denoising.

Accurate and fast image segmentation algorithms are of paramount importance for a wide range of medical imaging applications. Level set algorithms based on narrow band implementation have been among the most widely used segmentation algorithms. However, the accuracy of standard level set algorithms is compromised by the fact that their evolution schemes deteriorate the signed distance level set functions required for accurate computation of normals and curvatures. The most common remedy is to use an ad-hoc reinitialization step to rebuild the signed distance function frequently. Meanwhile, complex upwind finite difference schemes are required for stable evolution. They together make the overall computation expensive. In this paper, we propose a novel fast narrow band distance preserving level set evolution algorithm that eliminates the need for both reinitialization and complex upwind finite difference schemes. This is achieved by incorporating into a variational level set formulation with a signed distance preserving term that regularizes the evolution. As a result, stable, accurate, fast evolution could be obtained using a simple finite difference scheme within a very narrow band, defined as the union of all 3times3 pixel blocks around the zero crossing pixels. Also, our method allows the use of larger time step to speed up the convergence while ensuring accurate result, as well as the use of more general and computational efficient initial level set functions rather than the signed distance functions required by standard level set methods. The proposed algorithm has been applied on both synthetic and real images of different modalities with promising results

Active contours or snakes have been extensively utilized in handling image segmentation and classification problems. In traditional active contour models, snake initialization is performed manually by users, and topological changes, such as splitting of the snake, can not be automatically handled. In this paper, we introduce a new method to solve the snake initialization and splitting problem, based on an area segmentation approach: the external force field is segmented first, and then the snake initialization and splitting can be automatically performed by using the segmented external force field. Such initialization and splitting produces multiple snakes, each of which is within the capture range associated to an object and evolved to the object boundary. The external force used in this paper is a gradient vector flow with an edge-preserving property (EPGVF), which can prevent the snakes from passing over weak boundaries. To segment the external force field, we represent it with a graph, and a graph-theory approach can be taken to determine the membership of each pixel. Experimental results establish the effectiveness of the proposed approach.

In this paper, we present a new variational formulation for geometric active contours that forces the level set function to be close to a signed distance function, and therefore completely eliminates the need of the costly re-initialization procedure. Our variational formulation consists of an internal energy term that penalizes the deviation of the level set function from a signed distance function, and an external energy term that drives the motion of the zero level set toward the desired image features, such as object boundaries. The resulting evolution of the level set function is the gradient flow that minimizes the overall energy functional. The proposed variational level set formulation has three main advantages over the traditional level set formulations. First, a significantly larger time step can be used for numerically solving the evolution partial differential equation, and therefore speeds up the curve evolution. Second, the level set function can be initialized with general functions that are more efficient to construct and easier to use in practice than the widely used signed distance function. Third, the level set evolution in our formulation can be easily implemented by simple finite difference scheme and is computationally more efficient. The proposed algorithm has been applied to both simulated and real images with promising results.