Ultrasound is the preferred early cancer screening modality due to non-ionizing radiation, cost-effectiveness, and real-time imaging, yet conventional diagnosis relies heavily on physician expertise, causing significant subjectivity and limited efficiency. Vision-Language Models (VLMs) show promise but lack ultrasound-specific knowledge and multi-organ generalization. We propose EchoVLM, the first open-source 10-billion-parameter ultrasound-tailored VLM with a Mixture-of-Experts (MoE) architecture. It is infused with knowledge across seven anatomical systems, trained on 208,941 clinical cases, 1.47 million ultrasound key-frame images, and over 100 diseases or imaging findings. Supporting clinical report generation, diagnosis prediction, and Visual Question Answering (VQA), it outperforms Qwen2-VL by 7.58 BLEU-1 and 3.45 ROUGE-1 points in report generation. This work shows substantial potential for establishing a general-purpose ultrasound VLM and lays a technical foundation for clinical translation. Source code and model weights are available at https://github.com/Asunatan/EchoVLM.

Fault tolerance to interference and interpretability are crucial for computer-aided diagnosis (CAD) systems. In clinical practice, deep learning models are often not accepted by doctors and patients due to their lack of diagnostic reasoning capabilities. Moreover, ultrasound imaging data contain significant noise and artifacts, making deep models vulnerable to interference, which frequently leads to misclassification. To address this issue, this study leverages fuzzy logic to achieve fault-tolerant perception of image content and employs a fuzzy trustworthy brain-inspired reasoning framework for diagnostic decision-making. In the fast-thinking module, we innovatively introduce the fuzzification of deep classification probabilities and construct a fault-tolerant perception set based on medical feature classification information. In the slow-thinking module, we establish a tensor decomposition-based knowledge graph, where medical feature information and membership degrees from the fault-tolerant perception set are jointly embedded to form a trustworthy factor matrix. The reasoning space reconstructed under this matrix guides the reasoning process from medical features toward the correct diagnostic outcome. To validate the effectiveness of the proposed model, we conducted diagnostic experiments on thyroid nodules and breast tumors, achieving AUC scores of 0.9549 and 0.9751, respectively. Experimental results demonstrate that our approach exhibits outstanding diagnostic performance on real-world ultrasound data.

For the diagnosis of liver cancer, conventional brightness mode (B-mode) can only provide morphological information. Multi-modal ultrasound, including shear-wave elastography (SWE) and contrast enhanced ultrasound (CEUS), can provide comprehensive diagnostic information on tumor microenvironment and tissue perfusion . The challenge is to effectively explore the multi-modal features of ultrasound. Besides, there are many fusion strategies currently available, but there is a lack of systematic comparative research on the various fusion strategies. In this study, we designed 'Lesions Pairing' to construct the dataset, addressing the challenge of small sample sizes in multi-modal learning. We then compared the effectiveness of different strategies and proposed hybrid-fusion strategies based on the combination of conventional layer-level fusion (i.e. early-fusion, mid-fusion and late-fusion), which can efficiently extract intra-/inter- modal information. Specifically, we first systematically compared different deep-learning-based fusion strategies for multi-modal ultrasound in the diagnosis of liver cancer. Secondly, based on the comparison results of a multimodal framework that integrates B-mode, SWE, CEUS ultrasound data, and clinical data simultaneously, we propose a hybrid-fusion strategies for the diagnosis of hepatocellular carcinoma and intrahepatic cholangiocarcinoma . The experimental results showed that the area under the curve of the early-late fusion strategy combined with clinical data was 0.9854, which was superior to other single mode and other fusion strategies, increasing by 13.8–25.88 % and 2.22 %-9.79 %, respectively.

Lesion localization and tracking are critical for accurate, automated medical imaging analysis. Contrast-enhanced ultrasound (CEUS) significantly enriches traditional B-mode ultrasound with contrast agents to provide high-resolution, real-time images of blood flow in tissues and organs. However, many trackers, designed primarily for natural RGB or B-mode ultrasound images, underutilize the extensive data from dual-screen enhanced images and fail to account for respiratory motion, thus facing challenges in achieving accurate target tracking. To address the existing challenges, we propose an adaptive-weighted dual mapping (ADMNet), an online tracking framework tailored for CEUS. Firstly, we introduced a novel Multimodal Atrous Attention Fusion (MAAF) module, innovatively designed to adapt the weightage between B-mode and enhanced images in dual-screen CEUS, reflecting the clinician's dynamic focus shifts between screens. Secondly, we proposed a Respiratory Motion Compensation (RMC) module to correct motion trajectory interferences due to respiratory motion, effectively leveraging temporal information. We utilized two newly established CEUS datasets, totaling 35,082 frames, to benchmark the ADMNet against various advanced B-mode ultrasound trackers. Our extensive experiments revealed that ADMNet achieves new state-of-the-art performance in CEUS tracking. Ablation studies and visualizations further underline the effectiveness of MAAF and RMC modules, demonstrating the promising potential of ADMNet in clinical CEUS tracing, thus providing novel research avenues in this field.

Atypical Hepatocellular Carcinoma (HCC) is very hard to distinguish from Focal Nodular Hyperplasia (FNH) in routine imaging. However little attention was paid to this problem. This paper proposes a novel liver tumor Computer-Aided Diagnostic (CAD) approach extracting spatio-temporal semantics for atypical HCC. With respect to useful diagnostic semantics, our model automatically calculates three types of semantic feature with equally down-sampled frames based on Contrast-Enhanced Ultrasound (CEUS). Thereafter, a Support Vector Machine (SVM) classifier is trained to make the final diagnosis. Compared with traditional methods for diagnosing HCC, the proposed model has the advantage of less computational complexity and being able to handle the atypical HCC cases. The experimental results show that our method obtained a pretty considerable performance and outperformed two traditional methods. According to the results, the average accuracy reaches 94.40%, recall rate 94.76%, F1-score value 94.62%, specificity 93.62% and sensitivity 94.76%, indicating good merit for automatically diagnosing atypical HCC cases.