Abstract Ship classification in synthetic aperture radar (SAR) imagery plays a critical role in many practical applications. However, existing semi-supervised frameworks still suffer from unstable pseudo-label quality and insufficient utilization of unlabeled data. In this article, we propose a dual-teacher network with an adaptive reliable bank (DTRB-Net), a semi-supervised learning framework that integrates a dual-teacher architecture with an adaptive reliable bank to address these challenges. The teacher network and its subnetwork collaboratively generate pseudo-label pairs that are then filtered through a two-stage selection strategy based on pseudo-label confidence and prediction discrepancy to ensure reliability. These high-quality pseudo-label pairs are stored and dynamically updated in a class-wise adaptive reliable bank, providing stable contrastive samples for the student network, and enabling more effective exploitation of unlabeled data. In addition, we design a new loss function that jointly leverages labeled and unlabeled data to enhance the student network’s feature learning capability. Extensive experiments on the FUSAR-Ship and OpenSARShip datasets show that DTRB-Net achieves superior accuracy on both three- and six-class ship classification tasks compared with existing methods, demonstrating the effectiveness and robustness of the proposed framework.

The advent of Whole Slide Imaging (WSI) has revolutionised digital pathology by enabling computational analysis of gigapixel-scale images. To handle their large size, most deep learning models divide WSIs into patches and apply Multiple Instance Learning (MIL) for slide-level classification. However, MIL models often depend on pre-trained feature extractors, resulting in domain gaps between natural and pathological images. Parameter-Efficient Fine-Tuning (PEFT) via visual prompting has emerged to bridge this gap with minimal overhead. Nevertheless, existing visual prompts are typically attached at the image level and tightly coupled with specific architectures such as CNNs or ViTs, limiting generalisability and scalability in WSI tasks. To overcome these limitations, we propose Slide-aware Deep Feature Prompt (S-DFP), a novel visual prompting method which derives task-specific information directly from feature embeddings and is initialised with slide-specific cues, thereby enhancing compatibility with diverse feature extractors and MIL frameworks. Experiments on four benchmark datasets, CAMELYON16, BRIGHT, TCGA-IDH, and UniToPath, demonstrate that S-DFP consistently boosts MIL model performance by 2–5% in AUC while introducing less than 0.02% additional parameters. Furthermore, when integrated with recent pathology foundation models, S-DFP yields additional performance gains. The code is publicly available at S-DFP .

Stain imputation in multiplex immunofluorescence (mIF) imaging addresses the challenge of missing or damaged biomarker channels by reconstructing target biomarker images from a limited set of available stains. This approach offers a faster and more efficient alternative to full-panel staining, enabling detailed analysis of the tumour microenvironment. Existing One-Step Inference Models (OSIMs), primarily based on generative adversarial networks (GAN) or autoencoders, often generate suboptimal images with significant artifacts or reduced signal intensity. These limitations impair visual interpretability and reliability of the downstream immunotherapy response assessment. The challenge is further amplified when imputing cytoplasmic biomarkers such as CD8 from commonly used stains such as DAPI, due to the limited spatial correlation and the inherently complex structure of cytoplasmic signals. To address these limitations, we propose a self-refining diffusion model, OS2CR-Diff, which utilises the results from OSIMs as additional conditional representations. Unlike prior studies that rely on a single or limited conditional inputs, OS2CR-Diff incorporates three conditional inputs: the OSIM-imputed target biomarker image, OSIM-imputed complementary biomarker images, and non-antibody-stained images. Furthermore, we propose a feature fusion module that employs a cross-gated attention mechanism to effectively integrate these inputs, enabling context-aware feature refinement and improving the quality and reliability of imputed biomarker images. We evaluated OS2CR-Diff for CD8 biomarker imputation on mIF images of melanoma tissues. Our method outperformed state-of-the-art methods, achieving a$\mathbf{7 3. 4 \%}$increase in the Structural Similarity Index Measure (SSIM), a 28.9 % gain in the Peak Signal-to-Noise Ratio (PSNR), a$\mathbf{6 1. 2 \%}$improvement in Mean Absolute Error (MAE), and significantly lower false positive rates compared to OSIM.

Universal, transferable whole-slide image (WSI) representations are central to computational pathology. Incorporating multiple markers (e.g., immunohistochemistry, IHC) alongside H&E enriches H&E-based features with diverse, biologically meaningful information. However, progress is limited by the scarcity of well-aligned multi-stain datasets. Inter-stain Misalignment shifts corresponding tissue across slides, hindering consistent patch-level features and degrading slide-level embeddings. To address this, we curated a slide-level aligned, five-stain dataset (H&E, HER2, KI67, ER, PGR) to enable paired H&E-IHC learning and robust cross-stain representation. Leveraging this dataset, we propose Cross-Stain Contrastive Learning (CSCL), a two-stage pretraining framework: a lightweight adapter trained with patch-wise contrastive alignment to improve the compatibility of H&E features with corresponding IHC-derived contextual cues; and slide-level representation learning with Multiple Instance Learning (MIL), which uses a cross-stain attention fusion module to integrate stain-specific patch features and a crossstain global alignment module to enforce consistency among slide-level embeddings across different stains. Experiments on cancer subtype classification, IHC biomarker status classification, and survival prediction, show consistent gains by yielding high-quality, transferable H&E slide-level representations. The code and data are available at: https://github.com/lily-zyz/CSCL.

Since the inception of BraTS challenge, a series of methods has been developed for brain tumor segmentation over the past years. Although these methods achieved promising results, they mostly focus on glioma segmentation, largely due to their relatively high incidence. These fully-supervised methods may not be applicable as they rely on abundant labeled data, which is intrinsically inaccessible for rarer types of brain tumors. Data-efficient transfer learning approaches like few-shot learning and domain adaptation assume full access to source data, which may not be feasible in real-life scenarios due to privacy and confidentiality concerns. In this work, we propose a new source-free few-shot learning framework for rarer brain tumor segmentation that adapts source model trained on gliomas to other less common brain tumors such as meningioma, metastasis and pediatric tumors with only a few labeled target data. The proposed framework follows a dual-branch prototypes learning structure that harmonize preservation of common knowledge from source class and learning new features from target. We show that our method gains a 6% increase in Dice score over representative source-free domain adaptation methods, and achieves comparable performance against its fully-supervised counterpart.

Few-shot semantic segmentation (FSS) in remote sensing imagery remains challenging due to complex scenes, large image sizes, frequent multi-class coexistence, and significant intra-class variation. These factors often lead to severe overfitting to base classes and hinder generalization to novel classes. To address these issues, we propose a Mamba-Enhanced Transformer framework with Selective Prototype Aggregation (SPA), which simultaneously suppresses base-class interference and enhances novel-class representation. Specifically, the Dynamic Feature Aggregation (DFA) module integrates features from a base-class learner and a meta-learner, effectively enhancing inter-class distinction and mitigating feature contamination in multi-class coexistence scenarios. We further introduce the Mamba-Enhanced Transformer (MET), which combines Mamba blocks and Transformer self-attention to capture long-range dependencies and multi-scale context while maintaining low computational complexity. Mamba blocks independently extract support and query features, preventing cross-branch interference and improving robustness. By selectively aggregating support prototypes with query self-support prototypes, SPA strengthens feature discriminability and leverages query-specific contextual cues to reduce intra-class variance. Our model demonstrates strong performance across multiple datasets, achieves 50.49% mIoU on the iSAID dataset and 30.82% on the LoveDA dataset, representing improvements of 3.95% and 3.89%, respectively, over the Base model. Ablation experiments further validate the superior performance of SPA.

When training on a long-tailed dataset, conventional learning algorithms tend to exhibit a bias towards classes with a larger sample size. Our investigation has revealed that this biased learning tendency originates from the model parameters, which are trained to disproportionately contribute to the classes characterised by their sample size (e.g., many, medium, and few classes). To balance the overall parameter contribution across all classes, we investigate the importance of each model parameter to the learning of different class groups, and propose a multistage parameter Decouple and Optimisation (DO) framework that decouples parameters into different groups with each group learning a specific portion of classes. To optimise the parameter learning, we apply different training objectives with a collaborative optimisation step to learn complementary information about each class group. Extensive experiments on long-tailed datasets, including CIFAR100, Places-LT, ImageNet-LT, and iNaturaList 2018, show that our framework achieves competitive performance compared to the state-of-the-art.

Medical image datasets are often imbalanced due to biases in data collection and limitations in acquiring data for rare conditions. Addressing class imbalance is crucial for developing reliable deep-learning algorithms capable of effectively handling all classes. Recent class imbalanced methods have investigated the effectiveness of self-supervised learning (SSL) and demonstrated that such learned features offer increased resilience to class imbalance issues and obtain much improved performances over other types of class imbalanced methods. However, existing SSL methods either lack end-to-end capabilities or require substantial memory resources, potentially resulting in sub-optimal features and classifiers and limiting their practical usage. Moreover, the conventional pooling operations (e.g., max-pooling, or average-pooling) tend to generate less discriminative features when datasets pose high inter-class similarities. To alleviate the above issues, in this study, we present a novel end-to-end self-supervised learning framework tailored for imbalanced medical image datasets. Our framework constitutes an adaptive contrastive loss that can dynamically adjust the model’s learning focus between feature learning and classifier learning and a feature aggregation mechanism based on Graph Neural Networks to further enhance feature discriminability. We evaluate the effectiveness of our framework on four medical datasets, and the experimental results highlight its superior performance in imbalanced image classification tasks.

Abstract Histopathological analysis requires a lot of clinical experience and time for pathologists. Artificial intelligence (AI) may have an important role in assisting pathologists and leading to more efficient and effective histopathological diagnoses. To address the challenge of requiring a large number of labelled images to train deep learning models in breast cancer histopathological image classification, a self‐training semi‐supervised learning method consisting three components is proposed: Firstly, a pre‐trained ResNet‐18 was used to extract features and generate pseudo‐labels for unlabelled data; secondly, a relational weight network based on the squeeze‐and‐excitation network (SENet) was trained to calculate the non‐linear distance metrices between labelled and unlabelled samples, in order to improve the accuracy of pseudo‐labelling; lastly, a consistency loss—maximum mean difference (MMD)—was added into the model to minimize the divergence between distributions of unlabelled and labelled samples. Extensive experiments were conducted on the open access BreakHis dataset. The proposed method outperformed the state‐of‐the‐art semi‐supervised methods at all tested annotated percentages (10–70%), and also achieved comparable performance with supervised methods at higher annotated percentages (50%, 70%).

Brain tumor can be a fatal disease in the world. With the aim of improving survival rates, many computerized algorithms have been proposed to assist the pathologists to make a diagnosis' using Whole Slide Pathology Images (WSI). Most methods focus on performing patch-level classification and aggregating the patch-level results to obtain the image classification. Since not all patches carry diagnostic information, it is thus important for our algorithm to recognize discriminative and non-discriminative patches. In this study, we propose an iterative patch labelling algorithm based on the Convolutional Neural Network (CNN), with a well-designed thresholding scheme, a training policy and a novel discriminative model architecture, to distinguish patches and use the discriminative ones to achieve WSI -classification. Our method is evaluated on the MICCAI 2015 Challenge Dataset, and shows a large improvement over the baseline approaches.

Multi-View neuroimaging retrieval and classification play an important role in computer-aided-diagnosis of brain disorders, as multi-view features could provide more insights of the disease pathology and potentially lead to more accurate diagnosis than single-view features. The large inter-feature and inter-subject variations make the multi-view neuroimaging analysis a challenging task. Many multi-view or multi-modal feature fusion methods have been proposed to reduce the impact of inter-feature variations in neuroimaging data. However, there is not much in-depth work focusing on the inter-subject variations. In this study, we propose a subject-centered multi-view feature fusion method for neuroimaging retrieval and classification based on the propagation graph fusion (PGF) algorithm. Two main advantages of the proposed method are: 1) it evaluates the query online and adaptively reshapes the connections between subjects according to the query; 2) it measures the affinity of the query to the subjects using the subject-centered affinity matrices, which can be easily combined and efficiently solved. Evaluated using a public accessible neuroimaging database, our algorithm outperforms the state-of-the-art methods in retrieval and achieves comparable performance in classification.

Medical content-based retrieval (MCBR) plays an important role in computer aided diagnosis and clinical decision support. Multi-modal imaging data have been increasingly used in MCBR, as they could provide more insights of the diseases and complement the deficiencies of single-modal data. However, it is very challenging to fuse data in different modalities since they have different physical fundamentals and large value range variations. In this study, we propose a novel Propagation Graph Fusion (PGF) framework for multi-modal medical data retrieval. PGF models the subjects' relationships in single modalities using the directed propagation graphs, and then fuses the graphs into a single graph by summing up the edge weights. Our proposed PGF method could reduce the large inter-modality and inter-subject variations, and can be solved efficiently using the PageRank algorithm. We test the proposed method on a public medical database with 331 subjects using features extracted from two imaging modalities, PET and MRI. The preliminary results show that our PGF method could enhance multi-modal retrieval and modestly outperform the state-of-the-art single-modal and multi-modal retrieval methods.

The multi-view/multi-modal features are commonly used in neuroimaging classification because they could provide complementary information to each other and thus result in better classification performance than single-view features. However, it is very challenging to effectively integrate such rich features, since straightforward concatenation or singleview spectral embedding methods rarely leads to physically meaningful integration. In this paper, we present a supervised multi-view/multi-modal spectral embedding method (SMSE) for neuroimaging classification. This method embeds the high dimensional multi-view features derived from multi-modal neuroimaging data into a low dimensional feature space and preserves the optimal local embeddings among different views. The proposed SMSE algorithm, validated using three groups of neuroimaging data, is able to achieve significant classification improvement over the state-of-the-art multi-view spectral embedding methods.

Relevance feedback (RF) allows users to be actively involved in the information retrieval process and has been widely used in various information retrieval tasks. While most existing RF methods in content-based image retrieval (CBIR) focus on visual features of individual images only, in this paper we formulate the relevance feedback process as an energy minimization problem. The energy function takes into account both the feature aspect of each image and the manifold structure among individual images. The solution of labelling images as relevant or irrelevant is obtained with the graph cuts method. As a result, our method enables flexibly partitioning the feature space and labelling of images and is capable of handling challenging scenarios (or queries). Experimental results demonstrate that our proposed method outperforms the popular RF methods.

Accurate neuroimaging feature extraction is essential for effective content-based management of the large neuroimaging databases, as well as achieving improved diagnosis. In this paper, we presented a multiscale and multi-orientation neuroimaging feature extraction algorithm with degenerative patterns for content-based 3D neuroimaging analysis and retrieval, based on the localized 3D Gabor wavelets. Our proposed approach was evaluated with 209 3D clinical neurological imaging studies and compared with the 3D discrete curvelet transform based method and the 3D spatial grey level co-occurrence matrices based method. The preliminary results suggested that our algorithm could support more reliable 3D neuroimaging retrieval.

The volume and complexity of neurological images have significantly increased, which leads to challenges in efficient data management and retrieval. In this paper, we developed a new content-based image retrieval framework with the localized multiscale Discrete Curvelet Transform (DCvT) features extracted from parametric neurological images. We also compared the performance of three different irregular-to-regular shape padding methods. 142 patient data with neurodegenerative disorders were used in the evaluation. The preliminary results show that our proposed framework supports fast neuroimaging retrieval, and the orthographic projection method can reduce the computational complexity and has a great potential to improve the retrieval for indefinite cases.

Functional neuroimaging has an important role in non-invasive diagnosis of neurodegenerative disorders. There are now large volumes of imaging data generated by functional imaging technologies and so there is a need to efficiently manage and retrieve these data. In this paper, we propose a new scheme for efficient 3D content-based neurological image retrieval. 3D pathology-centric masks were adaptively designed and applied for extracting CMRGlc (cerebral metabolic rate of glucose consumption) texture features with volumetric co-occurrence matrices from neurological FDG PET images. Our results, using 93 clinical dementia studies, show that our approach offers a robust and efficient retrieval mechanism for relevant clinical cases and provides advantages in image data analysis and management.