A simultaneous enhancement of accuracy and diversity of predictions remains a challenge in ambiguous medical image segmentation (AMIS) due to the inherent trade-offs. While truncated diffusion probabilistic models (TDPMs) hold strong potential with a paradigm optimization, existing TDPMs suffer from entangled accuracy and diversity of predictions with insufficient fidelity and plausibility. To address the aforementioned challenges, we propose Ambiguity-aware Truncated Flow Matching (ATFM), which introduces a novel inference paradigm and dedicated model components. Firstly, we propose Data-Hierarchical Inference, a redefinition of AMIS-specific inference paradigm, which enhances accuracy and diversity at data-distribution and data-sample level, respectively, for an effective disentanglement. Secondly, Gaussian Truncation Representation (GTR) is introduced to enhance both fidelity of predictions and reliability of truncation distribution, by explicitly modeling it as a Gaussian distribution at Ttrunc instead of using sampling-based approximations. Thirdly, Segmentation Flow Matching (SFM) is proposed to enhance the plausibility of diverse predictions by extending semantic-aware flow transformation in Flow Matching (FM). Comprehensive evaluations on LIDC and ISIC3 datasets demonstrate that ATFM outperforms SOTA methods and simultaneously achieves a more efficient inference. ATFM improves GED and HM-IoU by up to 12% and 7.3% compared to advanced methods.

Predicting molecular properties is vital for drug discovery, but experimental measurement is costly and limited by scarce labeled data. Self-supervised molecular pretraining can leverage large unlabeled datasets, reducing dependence on extensive annotations. However, most methods struggle to preserve domain-specific chemical knowledge, especially clinically relevant substructures such as motifs. Random masking and generic graph augmentations often degrade critical chemical information and harm interpretability. Many approaches also work at a single scale-either atom or motif-missing opportunities for cross-scale integration. We propose A2M-Mol, a multi-perspective molecular pretraining framework that combines atom-level and motif-level views through four parallel graph constructions. This design enables cross-view alignment and multiscale fusion, explicitly encoding chemical knowledge. A2M-Mol employs a suite of self-supervised tasks, including cross-view correspondence, atomic reconstruction, global topology modeling, and property constraint enforcement, all coordinated via tailored contrastive learning. Extensive experiments across benchmarks and backbone architectures show consistent improvements over state-of-the-art methods. Ablation studies confirm strong synergies among the tasks. A2M-Mol maintains robust predictive accuracy across data scales, demonstrating effectiveness for real-world molecular property prediction and potential to accelerate drug discovery.

In domain continual medical image segmentation, distillation-based methods mitigate catastrophic forgetting by continuously reviewing old knowledge. However, these approaches often exhibit biases towards both new and old knowledge simultaneously due to confounding factors, which can undermine segmentation performance. To address these biases, we propose the Causality-Adjusted Data Augmentation (CauAug) framework, introducing a novel causal intervention strategy called the Texture-Domain Adjustment Hybrid-Scheme (TDAHS) alongside two causality-targeted data augmentation approaches: the Cross Kernel Network (CKNet) and the Fourier Transformer Generator (FTGen). (1) TDAHS establishes a domain-continual causal model that accounts for two types of knowledge biases by identifying irrelevant local textures (L) and domain-specific features (D) as confounders. It introduces a hybrid causal intervention that combines traditional confounder elimination with a proposed replacement approach to better adapt to domain shifts, thereby promoting causal segmentation. (2) CKNet eliminates confounder L to reduce biases in new knowledge absorption. It decreases reliance on local textures in input images, forcing the model to focus on relevant anatomical structures and thus improving generalization. (3) FTGen causally intervenes on confounder D by selectively replacing it to alleviate biases that impact old knowledge retention. It restores domain-specific features in images, aiding in the comprehensive distillation of old knowledge. Our experiments show that CauAug significantly mitigates catastrophic forgetting and surpasses existing methods in various medical image segmentation tasks.

Knowledge distillation can mitigate catastrophic forgetting in domain continual segmentation by transferring knowledge from the older model to the newer model. However, existing distillation-based methods primarily emphasize knowledge retention while overlooking inherent defects in the older teacher models. As a result, these teacher-originated defects, such as knowledge gaps or biases, are propagated and exacerbate forgetting. To address this challenge, we propose a Targeted Knowledge Rectification Learning framework (TKRL) to probe and correct teacher-originated defects. TKRL consists of two modules: 1) Probe-augmented Class Distillation, which generates gradient-driven "probes" to uncover underrepresented features in the older model, thereby bridging knowledge gaps by distilling hidden information into the new model; 2) Variance-guided Masked Autoencoder, which selectively masks and reconstructs critical high-uncertainty patches across multi-level semantic regions, thereby correcting biases inherited from the older model. Our experimental results show that TKRL effectively rectifies knowledge gaps and biases, thereby mitigating catastrophic forgetting and enhancing performance in domain continual segmentation.

T-cell receptors (TCR), which are heterodimers of $\alpha $ and $\beta $ chains that recognize foreign antigens, are of great significance to current immunotherapy. Although artificial intelligence (AI) has explosively accelerated de novo protein design, the challenge of therapeutic TCR design has been overlooked by most researchers. Existing TCR engineering relies heavily on isolating antigen-specific TCRs from tumor tissues, which requires a large amount of labor resources and wet experimental verification. To mitigate this issue, we present TCRdesign, a pretrained generative protein language model (PLM) for the de novo design of artificial TCR $\beta $-chain complementarity-determining region 3 sequences conditioned on antigen-binding specificity (BS). In parallel, we develop a high-accuracy binding predictor (TCRBinder) that couples paired $\alpha $/$\beta $ chain information with antigen sequences to assess BS. Our in silico comparisons demonstrate that (i) TCRdesign surpasses state-of-the-art baselines in generating antigen-specific TCR sequences. The model leverages paired-chain coherence to refine amino-acid level interaction patterns. (ii) TCRdesign-generated TCR sequences exhibit better antigen binding capability to diverse oncogenic hotspots compared with natural counterparts. (iii) TCRdesign inherits the intrinsic properties of large PLMs, enabling effectively identify the determinant residues in TCR-antigen binding, which enhances its interpretability. These results highlight the significant capability of TCRdesign in understanding and generating TCR sequences with an antigen-specific interaction pattern, charting a versatile path toward AI-driven T-cell engineering for precision immunotherapy.

Accurate prediction of binding between human leukocyte antigen (HLA) class I molecules and antigenic peptide segments is a challenging task and a key bottleneck in personalized immunotherapy for cancer. Although existing prediction tools have demonstrated significant results using established datasets, most can only predict the binding affinity of antigenic peptides to HLA and do not enable the immunogenic interpretation of new antigenic epitopes. This limitation results from the training data for the computational models relying heavily on a large amount of peptide-HLA (pHLA) eluting ligand data, in which most of the candidate epitopes lack immunogenicity. Here, we propose an adaptive immunogenicity prediction model, named MHLAPre, which is trained on the large-scale MS-derived HLA I eluted ligandome (mostly presented by epitopes) that are immunogenic. Allele-specific and pan-allelic prediction models are also provided for endogenous peptide presentation. Using a meta-learning strategy, MHLAPre rapidly assessed HLA class I peptide affinities across the whole pHLA pairs and accurately identified tumor-associated endogenous antigens. During the process of adaptive immune response of T-cells, pHLA-specific binding in the antigen presentation is only a pre-task for CD8+ T-cell recognition. The key factor in activating the immune response is the interaction between pHLA complexes and T-cell receptors (TCRs). Therefore, we performed transfer learning on the pHLA model using the pHLA-TCR dataset. In pHLA binding task, MHLAPre demonstrated significant improvement in identifying neoepitope immunogenicity compared with five state-of-the-art models, proving its effectiveness and robustness. After transfer learning of the pHLA-TCR data, MHLAPre also exhibited relatively superior performance in revealing the mechanism of immunotherapy. MHLAPre is a powerful tool to identify neoepitopes that can interact with TCR and induce immune responses. We believe that the proposed method will greatly contribute to clinical immunotherapy, such as anti-tumor immunity, tumor-specific T-cell engineering, and personalized tumor vaccine.

Adaptive immunity is a targeted immune response that enables the body to identify and eliminate foreign pathogens, playing a critical role in the anti-tumor immune response. Tumor cell expression of antigens forms the foundation for inducing this adaptive response. However, the human leukocyte antigens (HLA)-restricted recognition of antigens by T-cell receptors (TCR) limits their ability to detect all neoantigens, with only a small subset capable of activating T-cells. Accurately predicting neoantigen binding to TCR is, therefore, crucial for assessing their immunogenic potential in clinical settings. We present THLANet, a deep learning model designed to predict the binding specificity of TCR to neoantigens presented by class I HLAs. THLANet employs evolutionary scale modeling-2 (ESM-2), replacing the traditional embedding methods to enhance sequence feature representation. Using scTCR-seq data, we obtained the TCR immune repertoire and constructed a TCR-pHLA binding database to validate THLANet's clinical potential. The model's performance was further evaluated using clinical cancer data across various cancer types. Additionally, by analyzing divided complementarity-determining region (CDR3) sequences and simulating alanine scanning of antigen sequences, we provided new insights into the 3D binding interactions of TCRs and antigens. Predicting TCR-neoantigen pairing remains a significant challenge in immunology, THLANet provides accurate predictions using only the TCR sequence (CDR3β), antigen sequence, and class I HLA, offering novel insights into TCR-antigen interactions.

Automatic acne severity grading is crucial for the accurate diagnosis and effective treatment of skin diseases. However, the acne severity grading process is often ambiguous due to the similar appearance of acne with close severity, making it challenging to achieve reliable acne severity grading. Following the idea of fuzzy logic for handling uncertainty in decision-making, we transforms the acne severity grading task into a fuzzy label learning (FLL) problem, and propose a novel adjacency-aware fuzzy label learning (AFLL) framework to handle uncertainties in this task. The AFLL framework makes four significant contributions, each demonstrated to be highly effective in extensive experiments. First, we introduce a novel adjacency-aware decision sequence generation method that enhances sequence tree construction by reducing bias and improving discriminative power. Second, we present a consistency-guided decision sequence prediction method that mitigates error propagation in hierarchical decision-making through a novel selective masking decision strategy. Third, our proposed sequential conjoint distribution loss innovatively captures the differences for both high and low fuzzy memberships across the entire fuzzy label set while modeling the internal temporal order among different acne severity labels with a cumulative distribution, leading to substantial improvements in FLL. Fourth, to the best of our knowledge, AFLL is the first approach to explicitly address the challenge of distinguishing adjacent categories in acne severity grading tasks. Experimental results on the public ACNE04 dataset demonstrate that AFLL significantly outperforms existing methods, establishing a new state-of-the-art in acne severity grading.

Medical vision-language pretraining (VLP) that leverages naturally-paired medical image-report data is crucial for medical image analysis. However, existing methods struggle to accurately characterize associations between images and diseases, leading to inaccurate or incomplete diagnostic results. In this work, we propose MedFILIP, a fine-grained VLP model, introduces medical image-specific knowledge through contrastive learning, specifically: 1) An information extractor based on a large language model is proposed to decouple comprehensive disease details from reports, which excels in extracting disease deals through flexible prompt engineering, thereby effectively reducing text complexity while retaining rich information at a tiny cost. 2) A knowledge injector is proposed to construct relationships between categories and visual attributes, which help the model to make judgments based on image features, and fosters knowledge extrapolation to unfamiliar disease categories. 3) A semantic similarity matrix based on fine-grained annotations is proposed, providing smoother, information-richer labels, thus allowing fine-grained image-text alignment. 4) We validate MedFILIP on numerous datasets, e.g., RSNA-Pneumonia, NIH ChestX-ray14, VinBigData, and COVID-19. For single-label, multi-label, and fine-grained classification, our model achieves state-of-the-art performance, the classification accuracy has increased by a maximum of 6.69%.

Cardiovascular diseases are a major global health challenge, with electrocardiography (ECG) being critical for diagnosis and monitoring. As artificial intelligence and automated ECG diagnostic technologies rapidly advance, the demand for large-scale ECG databases continues to grow. Generative ECG has become a mainstream method to enhance database size and diversity. However, existing methods typically generate ECG randomly or focus on limited physiological categories, lacking the ability to synthesize ECG with varying physiological features and cardiac cycles, which is crucial for various practical applications. In response to this need, we propose a novel approach introducing a diffusion model called DIFF-ECG to generate precisely customized ECG that accurately reflect diverse cardiac conditions. Segmentation-based quality assessments confirmed that the synthesized ECG accurately followed the specified cardiac cycle information, with our model significantly outperforming baseline diffusion and GAN-based methods. Therefore, our approach addresses the critical need for generating clinically relevant and customizable ECG, contributing significantly to the field of automated cardiac disease diagnosis. By enabling fine-tuning of cardiac cycle phases, our method significantly expands the application range of generative ECG, potentially improving the diagnostic accuracy for rare diseases and advancing personalized medicine.

Faced with an escalating number of fingerprint images, most existing retrieval approachs suffer from a common problem: diminishing computational efficiency. This paper presents a hierarchical retrieval system tailored for high-resolution fingerprint images that utilizes abundant pore features and robust recognizability to improve retrieval performance. The framework comprises two core components. Firstly, a CNN-based feature extraction network is established, incorporating an attention mechanism to capture pore features in fingerprint images comprehensively. Subsequently, a hierarchical fingerprint retrieval approach is introduced, involving connection graph construction and a hierarchy of jump table structures for efficient retrieval of query pores. Empirical experiments conducted on high-resolution fingerprint image datasets underscore the system’s effectiveness. Compared with other advanced pore-based fingerprint retrieval methods, the proposed method exhibits a notable rise in the hit rate with reduced penetration rates, significantly reducing the retrieval time.

While significant strides have been made in predicting neoepitopes that trigger autologous CD4+ T cell responses, accurately identifying the antigen presentation by human leukocyte antigen (HLA) class II molecules remains a challenge. This identification is critical for developing vaccines and cancer immunotherapies. Current prediction methods are limited, primarily due to a lack of high-quality training epitope datasets and algorithmic constraints. To predict the exogenous HLA class II-restricted peptides across most of the human population, we utilized the mass spectrometry data to profile >223 000 eluted ligands over HLA-DR, -DQ, and -DP alleles. Here, by integrating these data with peptide processing and gene expression, we introduce HLAIImaster, an attention-based deep learning framework with adaptive domain knowledge for predicting neoepitope immunogenicity. Leveraging diverse biological characteristics and our enhanced deep learning framework, HLAIImaster is significantly improved against existing tools in terms of positive predictive value across various neoantigen studies. Robust domain knowledge learning accurately identifies neoepitope immunogenicity, bridging the gap between neoantigen biology and the clinical setting and paving the way for future neoantigen-based therapies to provide greater clinical benefit. In summary, we present a comprehensive exploitation of the immunogenic neoepitope repertoire of cancers, facilitating the effective development of "just-in-time" personalized vaccines.

MOTIVATION: Understanding the intermolecular interactions of ligand-target pairs is key to guiding the optimization of drug research on cancers, which can greatly mitigate overburden workloads for wet labs. Several improved computational methods have been introduced and exhibit promising performance for these identification tasks, but some pitfalls restrict their practical applications: (i) first, existing methods do not sufficiently consider how multigranular molecule representations influence interaction patterns between proteins and compounds; and (ii) second, existing methods seldom explicitly model the binding sites when an interaction occurs to enable better prediction and interpretation, which may lead to unexpected obstacles to biological researchers. RESULTS: To address these issues, we here present DrugMGR, a deep multigranular drug representation model capable of predicting binding affinities and regions for each ligand-target pair. We conduct consistent experiments on three benchmark datasets using existing methods and introduce a new specific dataset to better validate the prediction of binding sites. For practical application, target-specific compound identification tasks are also carried out to validate the capability of real-world compound screen. Moreover, the visualization of some practical interaction scenarios provides interpretable insights from the results of the predictions. The proposed DrugMGR achieves excellent overall performance in these datasets, exhibiting its advantages and merits against state-of-the-art methods. Thus, the downstream task of DrugMGR can be fine-tuned for identifying the potential compounds that target proteins for clinical treatment. AVAILABILITY AND IMPLEMENTATION: https://github.com/lixiaokun2020/DrugMGR.

The application of transformer architecture in analyzing whole slide images (WSIs) has become increasingly popular due to its remarkable ability to learn complex associations. Nevertheless, a significant drawback emerges in the multiclass analysis of WSIs. The majority of the transformer-based methods available currently rely primarily on a single, class-agnostic token. This approach might not ideally capture the subtleties of class-discriminative information. To address this challenge, we present an innovative approach tailored for multi-class WSI analysis that harnesses the power of class-specific tokens. Central to our method is a novel attention mechanism designed to foster a synergistic learning relationship between patch and class tokens, enhancing the granularity of information captured and ensuring a more comprehensive representation of the WSI. Complementing this, we introduce a dynamic class-centric training strategy designed to optimize token representation learning, ensuring each token is informatively aligned with its corresponding class. Through extensive experimentation on three challenging multi-class WSI analysis datasets, our method consistently demonstrates superior performance, underscoring its potential as a robust solution for multi-class WSI analysis tasks.

Breast cancer continues to be a pressing global health concern, emphasizing the essential need for effective diagnostic techniques. Automated Breast Ultrasound Systems (ABUS) provide a promising advance in breast tumor detection, yet they require significant expertise in interpreting 3D ABUS images, a task fraught with distinctive challenges. Although Vision Transformers (ViT) display remarkable potential for image processing, their low inductive bias and significant data requirements pose obstacles, particularly in the data-constrained medical field. To mitigate these issues, we introduce a Mask-Recover strategy for pretraining Transformer models on 3D ABUS images, enhancing model adaptability and reducing the data demands of the ViT model. Moreover, recognizing the risk that ViTs’ average pooling approach may unintentionally mask small but vital features, we propose Dual-CapsViT, an inventive model combining Transformers and Capsule Networks. This integration affords efficient token routing while preserving fine-grained details. To reconcile potential inconsistencies between capsules and tokens, we engineer a novel dual-channel routing algorithm, strengthening the decoder’s performance. We benchmarked our models against well-known standards such as ResNet and ViT for classifying breast tumors in ABUS images. Our models exhibited superior performance, as evidenced by improved accuracy, specificity, and Area Under the Receiver Operating Characteristic Curve (AUC) metrics, thereby affirming Dual-CapsViT’s potential to enhance breast cancer diagnostics.

Numerous researches emphasized that long non-coding RNA (lncRNA) plays a vital factor in various biological processes, and its mismatched expression and dysfunction are tightly linked with the occurrence of human diseases. Thus, computational models were designed to identify lncRNA-disease interactions by merging heterogeneous biological data. However, most of them neglected the intrinsic structure of multi-source information, which limits the performance for potential lncRNA-disease association prediction. Here, GCN-NIMC is introduced to alleviate the dilemma for disease-associated lncRNA genes identification based on the graph convolutional network with neural inductive matrix. This method builds a feature matrix with multi-source heterogeneous data and then learn the various information contained in the feature matrix for the sake of acquiring better feature expressions of the lncRNA-disease interactions. Experimental results on 10-repeated 5-fold cross-validation demonstrated that our proposed GCN-NIMC is superior to existing cutting-edge methods for identifying disease-related lncRNA genes. Furthermore, case studies confirmed our computational method as a practical tool with clinical benefits to develop the therapeutic schedule at lncRNA-level.

In this paper, we proposed a novel left ventricular volumes prediction method. This method is a cascade architecture which is based on multi-scale LV atlas location and deep convolutional neural networks (CNN). Firstly, we adopted LV atlas mapping method to achieve accurate location of LV region in cardiac magnetic resonance (CMR) images. And then, the CNN were used to train an end-to-end LV volumes prediction model to achieve the direct prediction. What's more, the large number of CMR images data (1140 subjects, more than 1026000 images) make the proposed deep CNN have relatively better feature representation and robust prediction ability. The experiment results on the large-scale CMR datasets prove that the proposed method has higher accuracy than the state-of-the-art prediction methods in terms of the end-diastole volumes (EDV), the end-systole volumes (ESV), and the ejection fraction (EF). Besides, we make the proposed method open accessible to public for wide application in other biomedical image processing fields.