Encoder-Decoder architectures are widely used in deep learning-based Deformable Image Registration (DIR), where the encoder extracts multi-scale features and the decoder predicts deformation fields by recovering spatial locations. However, current methods lack specialized extraction of features (that are useful for registration) and predict deformation jointly and homogeneously in all three directions. In this paper, we propose a novel expert-guided DIR network with Mixture of Experts (MoE) mechanism applied in both encoder and decoder, named SHMoAReg. Specifically, we incorporate Mixture of Attention heads (MoA) into encoder layers, while Spatial Heterogeneous Mixture of Experts (SHMoE) into the decoder layers. The MoA enhances the specialization of feature extraction by dynamically selecting the optimal combination of attention heads for each image token. Meanwhile, the SHMoE predicts deformation fields heterogeneously in three directions for each voxel using experts with varying kernel sizes. Extensive experiments conducted on two publicly available datasets show consistent improvements over various methods, with a notable increase from 60.58% to 65.58% in Dice score for the abdominal CT dataset. Furthermore, SHMoAReg enhances model interpretability by differentiating experts' utilities across/within different resolution layers. To the best of our knowledge, we are the first to introduce MoE mechanism into DIR tasks. The code will be released soon.

In this work, we leverage informative embeddings from foundational models for unsupervised anomaly detection in medical imaging. For small datasets, a memory-bank of normative features can directly be used for anomaly detection which has been demonstrated recently. However, this is unsuitable for large medical datasets as the computational burden increases substantially. Therefore, we propose to model the distribution of normative DINOv2 embeddings with a Dirichlet Process Mixture model (DPMM), a non-parametric mixture model that automatically adjusts the number of mixture components to the data at hand. Rather than using a memory bank, we use the similarity between the component centers and the embeddings as anomaly score function to create a coarse anomaly segmentation mask. Our experiments show that through DPMM embeddings of DINOv2, despite being trained on natural images, achieve very competitive anomaly detection performance on medical imaging benchmarks and can do this while at least halving the computation time at inference. Our analysis further indicates that normalized DINOv2 embeddings are generally more aligned with anatomical structures than unnormalized features, even in the presence of anomalies, making them great representations for anomaly detection. The code is available at https://github.com/NicoSchulthess/anomalydino-dpmm.

Both local details and global context are crucial in medical image segmentation, and effectively integrating them is essential for achieving high accuracy. However, existing mainstream methods based on CNN-Transformer hybrid architectures typically employ simple feature fusion techniques such as serial stacking, endpoint concatenation, or pointwise addition, which struggle to address the inconsistencies between features and are prone to information conflict and loss. To address the aforementioned challenges, we innovatively propose HiPerformer. The encoder of HiPerformer employs a novel modular hierarchical architecture that dynamically fuses multi-source features in parallel, enabling layer-wise deep integration of heterogeneous information. The modular hierarchical design not only retains the independent modeling capability of each branch in the encoder, but also ensures sufficient information transfer between layers, effectively avoiding the degradation of features and information loss that come with traditional stacking methods. Furthermore, we design a Local-Global Feature Fusion (LGFF) module to achieve precise and efficient integration of local details and global semantic information, effectively alleviating the feature inconsistency problem and resulting in a more comprehensive feature representation. To further enhance multi-scale feature representation capabilities and suppress noise interference, we also propose a Progressive Pyramid Aggregation (PPA) module to replace traditional skip connections. Experiments on eleven public datasets demonstrate that the proposed method outperforms existing segmentation techniques, demonstrating higher segmentation accuracy and robustness. The code is available at https://github.com/xzphappy/HiPerformer.

Achieving high image quality for temporal frames in dynamic positron emission tomography (PET) is challenging due to the limited statistic especially for the short frames. Recent studies have shown that deep learning (DL) is useful in a wide range of medical image denoising tasks. In this paper, we propose a model-based neural network for dynamic PET image denoising. The inter-frame spatial correlation and intra-frame structural consistency in dynamic PET are used to establish the kernel space-based multidimensional sparse (KMDS) model. We then substitute the inherent forms of the parameter estimation with neural networks to enable adaptive parameters optimization, forming the end-to-end neural KMDS-Net. Extensive experimental results from simulated and real data demonstrate that the neural KMDS-Net exhibits strong denoising performance for dynamic PET, outperforming previous baseline methods. The proposed method may be used to effectively achieve high temporal and spatial resolution for dynamic PET. Our source code is available at https://github.com/Kuangxd/Neural-KMDS-Net/tree/main.

Medical imaging provides critical evidence for clinical diagnosis, treatment planning, and surgical decisions, yet most existing imaging models are narrowly focused and require multiple specialized networks, limiting their generalization. Although large-scale language and multimodal models exhibit strong reasoning and multi-task capabilities, real-world clinical applications demand precise visual grounding, multimodal integration, and chain-of-thought reasoning. We introduce Citrus-V, a multimodal medical foundation model that combines image analysis with textual reasoning. The model integrates detection, segmentation, and multimodal chain-of-thought reasoning, enabling pixel-level lesion localization, structured report generation, and physician-like diagnostic inference in a single framework. We propose a novel multimodal training approach and release a curated open-source data suite covering reasoning, detection, segmentation, and document understanding tasks. Evaluations demonstrate that Citrus-V outperforms existing open-source medical models and expert-level imaging systems across multiple benchmarks, delivering a unified pipeline from visual grounding to clinical reasoning and supporting precise lesion quantification, automated reporting, and reliable second opinions.

Medical imaging provides critical evidence for clinical diagnosis, treatment planning, and surgical decisions, yet most existing imaging models are narrowly focused and require multiple specialized networks, limiting their generalization. Although large-scale language and multimodal models exhibit strong reasoning and multi-task capabilities, real-world clinical applications demand precise visual grounding, multimodal integration, and chain-of-thought reasoning. We introduce Citrus-V, a multimodal medical foundation model that combines image analysis with textual reasoning. The model integrates detection, segmentation, and multimodal chain-of-thought reasoning, enabling pixel-level lesion localization, structured report generation, and physician-like diagnostic inference in a single framework. We propose a novel multimodal training approach and release a curated open-source data suite covering reasoning, detection, segmentation, and document understanding tasks. Evaluations demonstrate that Citrus-V outperforms existing open-source medical models and expert-level imaging systems across multiple benchmarks, delivering a unified pipeline from visual grounding to clinical reasoning and supporting precise lesion quantification, automated reporting, and reliable second opinions.

Although deep learning models have greatly automated medical image segmentation, they still struggle with complex samples, especially those with irregular shapes, notable scale variations, or blurred boundaries. One key reason for this is that existing methods often overlook the importance of identifying and enhancing the instructive features tailored to various targets, thereby impeding optimal feature extraction and transmission. To address these issues, we propose two innovative modules: an Instructive Feature Enhancement Module (IFEM) and an Instructive Feature Integration Module (IFIM). IFEM synergistically captures rich detailed information and local contextual cues within a unified convolutional module through flexible resolution scaling and extensive information interplay, thereby enhancing the network's feature extraction capabilities. Meanwhile, IFIM explicitly guides the fusion of encoding-decoding features to create more discriminative representations through sensitive intermediate predictions and omnipresent attention operations, thus refining contextual feature transmission. These two modules can be seamlessly integrated into existing segmentation frameworks, significantly boosting their performance. Furthermore, to achieve superior performance with substantially reduced computational demands, we develop an effective and efficient segmentation framework (EESF). Unlike traditional U-Nets, EESF adopts a shallower and wider asymmetric architecture, achieving a better balance between fine-grained information retention and high-order semantic abstraction with minimal learning parameters. Ultimately, by incorporating IFEM and IFIM into EESF, we construct EE-Net, a high-performance and low-resource segmentation network. Extensive experiments across six diverse segmentation tasks consistently demonstrate that EE-Net outperforms a wide range of competing methods in terms of segmentation performance, computational efficiency, and learning ability. The code is available at https://github.com/duweidai/EE-Net.

Risk-adapted breast cancer screening requires robust models that leverage longitudinal imaging data. Most current deep learning models use single or limited prior mammograms and lack adaptation for real-world settings marked by imbalanced outcome distribution and heterogeneous follow-up. We developed LongiMam, an end-to-end deep learning model that integrates both current and up to four prior mammograms. LongiMam combines a convolutional and a recurrent neural network to capture spatial and temporal patterns predictive of breast cancer. The model was trained and evaluated using a large, population-based screening dataset with disproportionate case-to-control ratio typical of clinical screening. Across several scenarios that varied in the number and composition of prior exams, LongiMam consistently improved prediction when prior mammograms were included. The addition of prior and current visits outperformed single-visit models, while priors alone performed less well, highlighting the importance of combining historical and recent information. Subgroup analyses confirmed the model's efficacy across key risk groups, including women with dense breasts and those aged 55 years or older. Moreover, the model performed best in women with observed changes in mammographic density over time. These findings demonstrate that longitudinal modeling enhances breast cancer prediction and support the use of repeated mammograms to refine risk stratification in screening programs. LongiMam is publicly available as open-source software.

Automated analysis of neonatal fundus images presents a uniquely intricate challenge in medical imaging. Existing methodologies predominantly focus on diagnosing abnormalities from individual images, often leading to inaccuracies due to the diverse and subtle nature of neonatal retinal features. Consequently, clinical standards frequently mandate the acquisition of retinal images from multiple angles to ensure the detection of minute lesions. To accommodate this, we propose leveraging multiple fundus images captured from various regions of the retina to comprehensively screen for a wide range of neonatal ocular pathologies. We employ Multiple Instance Learning (MIL) for this task, and introduce a simple yet effective learnable structure on the existing MIL method, called Learnable Dense to Global (LD2G-MIL). Different from other methods that focus on instance-to-bag feature aggregation, the proposed method focuses on generating better instance-level representations that are co-optimized with downstream MIL targets in a learnable way. Additionally, it incorporates a bag prior-based similarity loss (BP loss) mechanism, leveraging prior knowledge to enhance performance in neonatal retinal screening. To validate the efficacy of our LD2G-MIL method, we compiled the Neonatal Fundus Images (NFI) dataset, an extensive collection comprising 115,621 retinal images from 8,886 neonatal clinical episodes. Empirical evaluations on this dataset demonstrate that our approach consistently outperforms stateof-the-art (SOTA) generic and specialized methods. The code and trained models are publicly available at https: //github.com/CVIU-CSU/LD2G-MIL.

Modern computer vision is converging on a closed loop in which perception, reasoning and generation mutually reinforce each other. However, this loop remains incomplete: the top-down influence of high-level reasoning on the foundational learning of low-level perceptual features is not yet underexplored. This paper addresses this gap by proposing a new paradigm for pretraining foundation models in downstream domains. We introduce Visual insTruction Pretraining (ViTP), a novel approach that directly leverages reasoning to enhance perception. ViTP embeds a Vision Transformer (ViT) backbone within a Vision-Language Model and pretrains it end-to-end using a rich corpus of visual instruction data curated from target downstream domains. ViTP is powered by our proposed Visual Robustness Learning (VRL), which compels the ViT to learn robust and domain-relevant features from a sparse set of visual tokens. Extensive experiments on 16 challenging remote sensing and medical imaging benchmarks demonstrate that ViTP establishes new state-of-the-art performance across a diverse range of downstream tasks. The code is available at github.com/zcablii/ViTP.