The use of imaging and genetic data for biomarker detection and disease diagnosis can deepen the understanding of disease pathogenesis and assist in clinical diagnosis. However, current methods face two major challenges: 1) the significant heterogeneity between multimodal data hampers modality fusion and 2) effectively exploring consistency and variability information from similar diseases for enhancing model performance is difficult. In this paper, we propose a novel unified framework, termed dual adaptive disentangled representation learning (DADRL), to simultaneously achieve disease-shared and disease-specific biomarker detection as well as disease diagnosis. Our DADRL comprises three components: 1) a biology information constraints-based modality fusion strategy is applied to adaptively explore inter- and intra-modal correlations, thereby effectively fusing multimodal data; 2) a unified framework that integrates modality fusion and disease diagnosis is proposed to mine disease-related information for simultaneously accomplishing disease-related biomarker detection and disease diagnosis; and 3) disentangled representation learning and several adaptive metric constraints are incorporated into the unified framework to adaptively separate disease-specific information from disease-shared feature representations for effectively identifying disease-shared and disease-specific biomarkers, thereby deepening the understanding of disease pathogenesis. Extensive experiments on multiple real datasets and simulated data demonstrate that our method significantly improves performance of biomarker detection and disease diagnosis.

The similar textures, diverse shapes and blurred boundaries of thyroid lesions in ultrasound images pose a significant challenge to accurate segmentation. Although several methods have been proposed to alleviate the aforementioned issues, their generalization is hindered by limited annotation data and insufficient ability to distinguish lesion from its surrounding tissues, especially in the presence of noise and outlier. Additionally, most existing methods lack uncertainty estimation which is essential for providing trustworthy results and identifying potential mispredictions. To this end, we propose knowledge-prompted trustworthy disentangled learning (KPTD) for thyroid ultrasound segmentation with limited annotations. The proposed method consists of three key components: 1) knowledge-aware prompt learning (KAPL) encodes TI-RADS reports into text features and introduces learnable prompts to extract contextual embeddings, which assist in generating region activation maps (serving as pseudo-labels for unlabeled images); 2) foreground-background disentangled learning (FBDL) leverages region activation maps to disentangle foreground and background representations, refining their prototype distributions through a contrastive learning strategy to enhance the model's discrimination and robustness; and 3) foreground-background trustworthy fusion (FBTF) integrates the foreground and background representations and estimates their uncertainty based on evidence theory, providing trustworthy segmentation results. Experimental results show that KPTD achieves superior segmentation performance under limited annotations, significantly outperforming state-of-the-art methods.

Automatic pancreas segmentation can facilitate diagnosis and treatment of pancreatic diseases. The combination of non-contrast, arterial, and venous phases of CT imaging can enhance differentiation of the pancreas from its surrounding structures. However, existing multimodal methods, which try to integrate the multimodal information in computer-aided pancreas segmentation, often overlook the inter-modal relationships and have a limited capability for information fusion. In this paper, we propose a multi-phase pancreas segmentation method for incorporating Feature Aggregation Module (FAM) and Modality Adaptive Transformer (MAT). Specifically, we use the venous phase as the primary modality, while the non-contrast and arterial phases serve as supplementary modalities, based on clinical prior knowledge. Our FAM integrates spatial information from the primary and supplementary modalities, while our MAT adaptively enhances feature representation and establishes long-range dependencies among modalities. Our method outperforms state-of-the-art techniques on a large scale dataset. Based on the segmented pancreas region, We further perform a downstream task focused on pancreatic volume calculation. The prediction accuracy is on par with manual segmentation, demonstrating effectiveness and potential application of our proposed method.

Automatic segmentation of pulmonary embolism (PE) in computed tomography pulmonary angiography (CTPA) facilitates the quantitative assessment of PE severity, which is crucial for accurate and comprehensive diagnosis and reducing the high mortality rate of PE. Recent studies have attempted to reduce segmentation errors by integrating vessel segmentation techniques. However, the PE segmentation performance of these methods is largely limited by inter-tissue similarities and the tiny size of PE, along with variability in the shape and position of PE. To address these issues, we propose a CT diagnostic mode-oriented and cross difficulty-aware network (DMCD-Net) for PE segmentation. Specifically, our DMCD-Net imitates the collaborative diagnostic mode of multi-modal CT to learn intensity differences between PE and surrounding tissues, which can effectively reduce false positive segmentation, especially in cases with tiny size and inter-tissue similarities. Moreover, we introduce a cross difficulty-aware scheme with cross-supervision strategies and a difficulty-aware loss function to enhance focus on difficult segmentation regions arising from the irregular shapes and variable locations of PE. Our DMCD-Net is evaluated on two different hospitals and two public datasets. Extensive experiments demonstrate that DMCD-Net outperforms the state-of-the-art methods and shows better generalizability in PE segmentation.

High pitch helical Computed Tomography (CT) scanning significantly reduces radiation dose while improving temporal resolution, offering substantial clinical benefits. However, the incomplete scanning data commonly leads to artifacts in the reconstructed images, degrading image quality and potentially affecting clinical diagnosis. Existing high pitch reconstruction methods primarily operate within the image domain or combine image-domain networks with traditional iterative algorithms, yet their performance remains limited. To address such limitations, we propose Delta-Net, a deep dual-domain alternating iterative optimization network for high pitch helical CT reconstruction. We introduce a novel optimization objective and develop an alternating iterative optimization framework, where each sub iteration consists of projection domain correction and image domain refinement. To enhance generalization and robustness, deep neural networks are employed to learn domain-specific priors, which are incorporated as regularization terms, with all hyper-parameters automatically optimized during training. Specifically, the image domain residual refinement network (IRN) and projection domain consistency enhanced network (PCN) regularize the intermediate results across both domains. Additionally, to improve the capability of artifact suppression and structure restoration, a structure-aware joint loss is tailored for the optimization of Delta-Net. Quantitative and qualitative evaluations on clinical datasets demonstrate that Delta-Net outperforms other competitive methods in artifact suppression, fine structure recovery, and generalization.

Noninvasively predicting the status of isocitrate dehydrogenase (IDH) and chromosome arms 1p/19q preoperatively on multisequence magnetic resonance imaging (MRI) images is helpful for prognosis and optimal therapy planning of patients with gliomas. However, effectively learning discriminative features from MRI images for predicting IDH mutation and 1p/19q codeletion status remains challenging due to the high heterogeneity of gliomas. A dual structural feature exploration and alignment network (DSFEAnet) was proposed to effectively explore representative features associated with the intratumoral and marginal heterogeneity of gliomas for accurate prediction. First, a match and mismatch feature extraction (MMFE) module was introduced to extract image structural features related to intratumoral heterogeneity, such as information associated with tumor core localization and T2-fluid-attenuated inversion recovery (FLAIR) mismatch sign. Second, a graph-based geometry exploration (GGE) module was developed to explore graph structural features related to marginal heterogeneity. In this module, the vertex associations and variations perpendicularly along a 3-D tumor surface were integrated as a graph, which can effectively perceive changes in locations, sizes, and marginal textures of gliomas, thus enhancing the feature representational ability to describe glioma heterogeneity. Finally, a dual structural feature alignment (DSFA) module was incorporated to narrow the gaps among intra- and interstructural features. It can adaptively align and fuse different features and thus further improve the overall prediction performance. The proposed DSFEAnet was evaluated using a multicenter dataset, and its robustness was demonstrated on an independent clinical dataset. Specifically, preoperative MRI images of 560 glioma samples were collected from publicly available The Cancer Imaging Archive (TCIA) ( $n =203$ , age: 51.82 (15.21) years, male/female: 109/94, IDH-mutant/wild-type: 90/113, 1p/19q-codeleted/noncodeleted: 27/176), Nanfang hospital ( $n =136$ , age: 41.96 (12.29) years, male/female: 80/56, IDH-mutant/wild-type: 53/83, 1p/19q-codeleted/noncodeleted: 31/105), and Zhujiang hospital ( $n =221$ , age: 43.82 (17.36) years, male/female: 134/87, IDH-mutant/wild-type: 94/127, 1p/19q-codeleted/noncodeleted: 34/187). Our DSFEAnet achieved an AUC of 87.72% for IDH mutation status prediction and an AUC of 80.52% for 1p/19q codeletion status prediction in the Nanfang hospital dataset. Finally, the interpretability of the proposed modules was assessed to highlight the effectiveness of our method. Overall, the DSFEAnet exhibits great potential for predicting IDH mutation and 1p/19q codeletion status.

Ultrasound image segmentation is critical for tumor screening but is hindered by noise, artifacts, and high variability in lesion appearance. Challenges like blurred boundaries and morphological similarities further complicate accurate delineation. To address this, we propose the Dynamic Edge-aware Transformer with Cross-image Patch Dependency (DET-CPD). Our model integrates two key modules: a Dynamic Difference Convolution Module (DDCM) to enhance edge representation for varied lesions, and a Cross-Scale Semantic Enhancement Module (CSEM) that leverages cross-scale channel information to distinguish tumors from surrounding tissue. Crucially, we introduce a novel Cross-image Patch Dependency Loss (CPDLoss) that captures semantic dependencies across different images in a batch, improving robustness. Extensive experiments on four public datasets (BUSI, DatasetB, DDTI, and TN3K) demonstrate that DET-CPD achieves state-of-the-art segmentation performance.

X-ray images play an important role in the diagnosis of spinal diseases because of their convenient collection and easy observation. But it is time-consuming and challenging for radiologists to examine the differences between the vertebrae to diagnose abnormalities and locate lesions. Many existing methods try to extract the global features of radiographs and do not make full use of adjacent vertebrae variations. In this paper, we propose a novel Axial-aware neural network with Consecutive Attention Transformer (CAT), namely ACAT-Net, which takes advantage of the convolutional neural network and transformer as a new deep learning framework. A deep convolutional network extracts features of anteroposterior and lateral X-ray images that may have abnormalities in them. The consecutive attention transformer block is then used to focus on the morphological differences of axial adjacent vertebrae on the spines. The ingenious structure we designed can significantly reduce the amount of network parameters. Extensive experiments on clinical and public datasets show that our method is remarkably superior to other existing approaches in the spine X-ray image analysis.

Noise detection accuracy is crucial in suppressing random-valued impulse noise. Both false and miss detections determine the final estimation performance. Deterministic detection methods, which distinctly classify pixels into noisy or uncorrupted pixels, tend to increase the estimation error because some uncorrupted edge points are hard to discriminate from the random-valued impulse noise points. This paper proposes an iterative structure-adaptive fuzzy estimation (SAFE) for random-valued impulse noise suppression. This SAFE method is developed in the framework of Gaussian maximum likelihood estimation. The structure-adaptive fuzziness is reflected by two structure-adaptive metrics based on pixel reliability and patch similarity, respectively. The reliability metric for each pixel (as noise free) is estimated via a novel-minimal-path-based structure propagation to give full consideration of the spatially varying image structures. A robust iteration stopping strategy is also proposed by evaluating the reestimation error of the uncorrupted intensity information. The comparative experimental results show that the proposed structure-adaptive fuzziness can lead to effective restoration. An efficient implementation of this SAFE method is also realized via graphics-processing-unit-based parallelization.