Ultrasound-CLIP: Semantic-Aware Contrastive Pre-training for Ultrasound Image-Text Understanding

• URL: http://arxiv.org/abs/2604.01749v1
• Date: Thu, 02 Apr 2026 08:10:31 GMT
• Title: Ultrasound-CLIP: Semantic-Aware Contrastive Pre-training for Ultrasound Image-Text Understanding
• Abstract summary: Existing vision-language pre-training models, such as CLIP, are primarily designed for other modalities.<n>We construct US-365K, a large-scale ultrasound image-text dataset containing 365k paired samples across 52 anatomical categories.<n>We propose Ultrasound-CLIP, a semantic-aware contrastive learning framework that introduces semantic soft labels and semantic loss to refine sample discrimination.
• Score: 14.606536256652726
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Ultrasound imaging is widely used in clinical diagnostics due to its real-time capability and radiation-free nature. However, existing vision-language pre-training models, such as CLIP, are primarily designed for other modalities, and are difficult to directly apply to ultrasound data, which exhibit heterogeneous anatomical structures and diverse diagnostic attributes. To bridge this gap, we construct US-365K, a large-scale ultrasound image-text dataset containing 365k paired samples across 52 anatomical categories. We establish Ultrasonographic Diagnostic Taxonomy (UDT) containing two hierarchical knowledge frameworks. Ultrasonographic Hierarchical Anatomical Taxonomy standardizes anatomical organization, and Ultrasonographic Diagnostic Attribute Framework formalizes nine diagnostic dimensions, including body system, organ, diagnosis, shape, margins, echogenicity, internal characteristics, posterior acoustic phenomena, and vascularity. Building upon these foundations, we propose Ultrasound-CLIP, a semantic-aware contrastive learning framework that introduces semantic soft labels and semantic loss to refine sample discrimination. Moreover, we construct a heterogeneous graph modality derived from UDAF's textual representations, enabling structured reasoning over lesion-attribute relations. Extensive experiments with patient-level data splitting demonstrate that our approach achieves state-of-the-art performance on classification and retrieval benchmarks, while also delivering strong generalization to zero-shot, linear probing, and fine-tuning tasks.

Related papers

• UltraSAM3: A Concept-Driven Foundation Model for Universal Ultrasound Image Segmentation [22.394292096035603]
  UltraSAM3 is a concept-driven foundation model for universal ultrasound image segmentation.<n>It is trained on a large-scale ultrasound segmentation corpus covering 37 public datasets and 13 anatomical categories.<n>Experiments demonstrate that UltraSAM3 consistently outperforms representative concept- and text-driven biomedical segmentation models.
  arXiv  Detail & Related papers  (2026-07-31T09:19:49Z)
• Echo-α: Large Agentic Multimodal Reasoning Model for Ultrasound Interpretation [76.6507710204181]
  We present Echo-, an agentic multimodal reasoning model for ultrasound interpretation.<n> Echo- is trained to coordinate organ-specific detector outputs, integrate them with global visual context, and convert the resulting evidence into grounded diagnostic decisions.<n>Results suggest that agentic multimodal reasoning can turn specialized detectors into verifiable clinical evidence.
  arXiv  Detail & Related papers  (2026-04-30T15:31:00Z)
• Clinical Cognition Alignment for Gastrointestinal Diagnosis with Multimodal LLMs [63.535652574541764]
  Multimodal Large Language Models (MLLMs) have demonstrated remarkable potential in medical image analysis.<n>Their application in gastrointestinal endoscopy is currently hindered by two critical limitations.<n>We propose a novel Clinical-Cognitive-Aligned (CogAlign) framework to address these challenges.
  arXiv  Detail & Related papers  (2026-03-21T07:47:37Z)
• A Fully Open and Generalizable Foundation Model for Ultrasound Clinical Applications [77.3888788549565]
  We present EchoCare, a novel ultrasound foundation model for generalist clinical use.<n>We developed EchoCare via self-supervised learning on our curated, publicly available, large-scale dataset EchoCareData.<n>With minimal training, EchoCare outperforms state-of-the-art comparison models across 10 representative ultrasound benchmarks.
  arXiv  Detail & Related papers  (2025-09-15T10:05:31Z)
• NeuroXVocal: Detection and Explanation of Alzheimer's Disease through Non-invasive Analysis of Picture-prompted Speech [4.815952991777717]
  NeuroXVocal is a novel dual-component system that classifies and explains potential Alzheimer's Disease (AD) cases through speech analysis.<n>The classification component (Neuro) processes three distinct data streams: acoustic features capturing speech patterns and voice characteristics, textual features extracted from speech transcriptions, and precomputed embeddings representing linguistic patterns.<n>The explainability component (XVocal) implements a Retrieval-Augmented Generation (RAG) approach, leveraging Large Language Models combined with a domain-specific knowledge base of AD research literature.
  arXiv  Detail & Related papers  (2025-02-14T12:09:49Z)
• Deep Spectral Methods for Unsupervised Ultrasound Image Interpretation [53.37499744840018]
  This paper proposes a novel unsupervised deep learning strategy tailored to ultrasound to obtain easily interpretable tissue separations. We integrate key concepts from unsupervised deep spectral methods, which combine spectral graph theory with deep learning methods. We utilize self-supervised transformer features for spectral clustering to generate meaningful segments based on ultrasound-specific metrics and shape and positional priors, ensuring semantic consistency across the dataset.
  arXiv  Detail & Related papers  (2024-08-04T14:30:14Z)
• CathFlow: Self-Supervised Segmentation of Catheters in Interventional Ultrasound Using Optical Flow and Transformers [66.15847237150909]
  We introduce a self-supervised deep learning architecture to segment catheters in longitudinal ultrasound images. The network architecture builds upon AiAReSeg, a segmentation transformer built with the Attention in Attention mechanism. We validated our model on a test dataset, consisting of unseen synthetic data and images collected from silicon aorta phantoms.
  arXiv  Detail & Related papers  (2024-03-21T15:13:36Z)
• HiFi-Syn: Hierarchical Granularity Discrimination for High-Fidelity Synthesis of MR Images with Structure Preservation [11.924728060845595]
  We introduce hierarchical granularity discrimination, which exploits various levels of semantic information present in medical images. Our strategy utilizes three levels of discrimination granularity: pixel-level discrimination using a Brain Memory Bank, structure-level discrimination on each brain structure with a re-weighting strategy to focus on hard samples. Our model may offer an alternative solution in scenarios where specific MR modalities of patients are unavailable.
  arXiv  Detail & Related papers  (2023-11-21T09:15:24Z)
• Factored Attention and Embedding for Unstructured-view Topic-related Ultrasound Report Generation [70.7778938191405]
  We propose a novel factored attention and embedding model (termed FAE-Gen) for the unstructured-view topic-related ultrasound report generation. The proposed FAE-Gen mainly consists of two modules, i.e., view-guided factored attention and topic-oriented factored embedding, which capture the homogeneous and heterogeneous morphological characteristic across different views.
  arXiv  Detail & Related papers  (2022-03-12T15:24:03Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.