Diaphragm ultrasound has become a widely adopted bedside tool for assessing respiratory muscle function in both acute and chronic respiratory failure. Initially introduced as a descriptive imaging technique, it is now increasingly used to support physiological interpretation, prognostic evaluation, and monitoring during invasive and noninvasive respiratory support, as well as in chronic respiratory diseases. In clinical practice, diaphragm ultrasound findings are often interpreted using isolated threshold values to classify diaphragm function as normal or abnormal, an approach that may inadequately represent the underlying respiratory muscle physiology. This state-of-the-art review summarizes current evidence across critical illness, noninvasive respiratory support, chronic respiratory failure, and chronic obstructive pulmonary disease (COPD). Conventional parameters—such as diaphragm thickness, thickening fraction, and excursion—are discussed within a physiological framework, highlighting their limitations in assessing intrinsic muscle strength when considered in isolation. Emerging ultrasound-based techniques, including speckle tracking, elastography, echogenicity analysis, contrast-enhanced ultrasound, and area-based motion assessment, are also reviewed. These approaches aim to provide deeper insights into diaphragmatic biomechanics, muscle quality, and perfusion, although they remain largely investigational and lack standardized acquisition and interpretation protocols. Overall, conventional diaphragm ultrasound parameters offer partial, load-dependent information with moderate and variable prognostic value across clinical settings. Integrated approaches combining diaphragm, lung, and peripheral muscle ultrasound appear to enhance physiological interpretation and clinical relevance. Diaphragm ultrasound should therefore be regarded as a dynamic, complementary component of multimodal respiratory assessment rather than a stand-alone diagnostic tool, with its greatest value in longitudinal monitoring and individualized clinical decision making.
PURPOSE:To develop and externally validate an ultrasound-based habitat subregional radiomics model for preoperative prediction of invasive breast cancer with concomitant ductal carcinoma in situ (IBC-DCIS). METHODS:A total of 1063 pathologically confirmed breast cancer patients from two centers were retrospectively enrolled and divided into a training cohort (n = 637) and an external validation cohort (n = 426). Tumor regions of interest were manually delineated on two-dimensional ultrasound images and further partitioned into three intratumoral habitat subregions using unsupervised clustering. Radiomics features were extracted from the whole tumor and each subregion. Feature selection was performed using Pearson correlation analysis and least absolute shrinkage and selection operator regression. Multiple machine learning models were constructed and evaluated using the area under the receiver operating characteristic curve (AUC) with 95% confidence intervals, calibration curves, and decision curve analysis. Model comparisons were conducted using the DeLong test. RESULTS:The support vector machine-based combined model achieved the highest AUC in the external validation cohort, with an AUC of 0.910 (95% CI: 0.883-0.936), and showed acceptable calibration and clinical net benefit across a broad range of threshold probabilities. DeLong test results showed that the combined model significantly outperformed imaging-based and single-region radiomics models (p < 0.05). To account for potential class imbalance, model performance was further assessed using multiple complementary metrics, including sensitivity, specificity, predictive values, balanced accuracy, F1-score, PR-AUC, and Brier score. CONCLUSION:An ultrasound-based habitat subregional radiomics model showed favorable performance for the preoperative prediction of IBC-DCIS and may provide supplementary information for preoperative risk stratification.
OBJECTIVE:Breast ultrasound imaging is widely used for the early detection of malignant breast lesions. Although deep learning models have shown strong performance, most approaches rely solely on raw image features and overlook clinically significant characteristics essential for reliable diagnosis. This study aims to develop a graph-based neural network model that integrates medically meaningful features to improve malignancy detection in breast ultrasound images. METHODS:The Breast Ultrasound (BUS) and Breast Ultrasound 2 (BUS2) datasets were pre-processed for region extraction using a bit-wise AND operation. Clinically relevant tumor features, such as shape, texture and statistical attributes, were extracted using Histogram of Oriented Gradients, gray-level co-occurrence matrix (GLCM) and histogram descriptors from the BUS, BUS2 and Mendeley Breast Ultrasound datasets. Each image was represented as a graph node with feature attributes, with the edges constructed using k-nearest neighbors based on feature similarity. A GraphSAGE-based graph neural network classifier with mean, max and long short-term memory (LSTM) aggregators was trained and evaluated. RESULTS:On the BUS dataset, the proposed approach achieved accuracies of 92.89% (mean), 93.20% (max) and 94.13% (LSTM). On the Mendeley Breast Ultrasound dataset, accuracies reached 99.80% (mean), 99.60% (max) and 99.99% (LSTM). For the BUS2 dataset, accuracies were 71.96% (mean), 73.15% (max) and 73.19% (LSTM). On the combined BUS-BUS2 dataset, the model achieved 86.69% (mean), 86.58% (max) and 87.62% (LSTM). CONCLUSION:The proposed GraphSAGE-based graph neural network effectively incorporates clinically significant features for breast ultrasound malignancy detection and demonstrates strong generalization across multiple datasets. By modeling feature-based relationships among ultrasound images through a graph structure, the proposed framework captures similarities between lesions and enhances classification performance. These findings highlight the potential of graph-based representations for improving computer-aided breast cancer diagnosis using medically meaningful descriptors.
OBJECTIVE:Ultrasound-mediated sonoporation has been explored for intracellular delivery in T cells, yet most studies rely on immortalized cell lines, limiting translational relevance. This study aimed to directly compare acoustofluidic molecular delivery efficiency and intracellular distribution in Jurkat T cells and primary human T cells. METHODS:Ultrasound-mediated molecular delivery was assessed using cationic microbubbles in a 3D-printed acoustofluidic device. Ultrasound-mediated delivery efficiency of 150 kDa fluorescein-labeled dextran and green fluorescent protein-expressing plasmid (2.9 MDa) were quantified using flow cytometry. Confocal microscopy was employed to evaluate intracellular distribution. RESULTS:Sonoporation enhanced dextran uptake in both cell types while maintaining high viability; however, Jurkat T cells exhibited significantly greater uptake than primary human T cells under matched conditions. Confocal imaging revealed broad intracellular distribution of dextran in both cell types, with comparable fluorescence intensity in cytoplasmic and nuclear regions. Within primary human T cells, activated cells demonstrated greater dextran uptake than non-activated cells, while no significant differences were observed between CD4⁺ and CD8⁺ subsets. In contrast to dextran delivery, plasmid delivery resulted in modest, non-significant increases in gene expression in both Jurkat and primary T cells. CONCLUSION:Acoustofluidic treatment enhanced intracellular delivery in both Jurkat and primary human T cells, while revealing important cell-type-dependent differences in delivery outcomes. These results reveal key differences in molecular delivery between immortalized T cell lines and primary human T cells, which provide important insights for the development of ultrasound-based delivery approaches to advance clinical T cell engineering technologies.
OBJECTIVE:Deep learning-based automated analysis of transgastric short-axis view (TSV) transesophageal echocardiography (TEE) remains under-explored. In this study, we propose a deep learning-based method for fully automated left ventricular segmentation and ejection fraction (EF) prediction in TSV TEE videos. METHODS:We built upon the U-Net network and proposed an Echo Efficient U-Net (Echo-EU-Net) segmentation model by replacing the original standard convolutions with depth-wise separable convolutions and by introducing the Multi-Efficient Channel Attention (MECA) and Enhanced Atrous Spatial Pyramid Pooling (EASPP) modules. We also incorporated automatic cardiac phase tracking and EF calculation. Experiments were performed on a TSV TEE dataset containing 694 videos from 451 patients, with expert manual segmentations and manual EF measurements as the reference standard. RESULTS:The proposed Echo-EU-Net, with an average Dice similarity coefficient of 92.91% and a Jaccard similarity coefficient of 87.23%, outperformed U-Net and its variants for left ventricular segmentation in TSV TEE, particularly in challenging cases. The model parameter size of Echo-EU-Net was 1.30 million, compared with 7.79 million for U-Net. The proposed EF prediction method had a satisfying agreement with the manual EF measurements (Pearson's r=0.84), with a mean absolute error of 6.44%. An ablation study demonstrated the effectiveness of the MECA and EASPP modules. CONCLUSION:The feasibility of the proposed Echo-EU-Net-based method in automatically segmenting the left ventricle and measuring EF in TSV TEE has been demonstrated. The findings of this study may shed light on lightweight deep learning-based fully automated left ventricular segmentation and EF quantification in TSV TEE.
OBJECTIVE:To evaluate complexity pursuit as an interpretable decomposition of lung ultrasound video dynamics and determine whether pleural-proximal spatial-temporal descriptors track ordinal severity. METHODS:A 46-video public cohort was used for construct validation across COVID-19, normal, bacterial pneumonia and viral pneumonia videos. Complexity-pursuit modes were represented by temporal traces and spatial maps and compared with principal components, random projection and spatial shuffles. After geometry correction and confound screening, four descriptors were retained. Severity validation used a separate, non-overlapping set of 151 COVIDx-US videos with dataset-encoded severity scores from 0 to 3. RESULTS:The fixed pleural-proximal energy fraction and its region-of-interest-normalized counterpart were moderately associated with severity (ρ=-0.576 and -0.572; both q<10-13). Associations remained after source adjustment (ρ=-0.355 and -0.365 ; both q<10-4 ) and preserved direction across all leave-one-source-out analyses. In proportional-odds models, each standard-deviation increase in either descriptor was associated with a lower odds of a higher severity score (odds ratios, 0.26 and 0.27). Adjacent-pair area under the curve point estimates ranged from 0.64 to 0.72, with greater uncertainty for comparisons involving the small score-1 sub-group. The two pleural descriptors were nearly collinear and did not support a combined multivariable score. CONCLUSION:Complexity-pursuit pleural-proximal topography provides a compact and interpretable representation of severity-linked organization in lung ultrasound videos.
OBJECTIVE:The suprascapular nerve (SSN) provides major motor and sensory innervation to the shoulder. Its accurate identification on ultrasound is challenging because of its small size, low contrast and proximity to structures with similar echotexture. This study aimed to develop and evaluate a deep learning-based approach for multi-class segmentation of the SSN and adjacent structures in dynamic supraclavicular ultrasound. METHODS:Dynamic ultrasound videos (n = 80) from 42 healthy adults were manually annotated for the SSN, brachial plexus, subclavian artery and omohyoid muscle. A Double U-Net architecture was implemented, incorporating a VGG-19 pre-trained encoder in the first stage and a randomly initialized encoder in the second stage, together with Atrous Spatial Pyramid Pooling and Squeeze-and-Excitation blocks. Ablation experiments examined the effects of encoder choice, architectural design, annotation strategy, loss function and output weighting. Segmentation performance was evaluated using Dice similarity coefficients on an independent test set. RESULTS:The Double U-Net significantly outperformed the baseline U-Net across all annotated structures. For SSN segmentation, the mean Dice coefficient improved from 0.51 ± 0.19 to 0.68 ± 0.11 (p = 0.03), with greater stability across test videos. A pre-trained VGG-19 encoder showed higher Dice scores than alternative encoders, without statistical significance (p = 0.12). Ablation analyses confirmed complementary contributions of both stages (p = 0.52). Multi-structure supervision yielded modest benefit, with optimal performance when all four structures were annotated (Dice 0.68 ± 0.11). Focal loss achieved the highest average Dice score, and optimal performance was obtained by emphasizing the final output with limited intermediate supervision (α:β = 0.1:0.9). CONCLUSION:The proposed Double U-Net enables reliable multi-class segmentation of the SSN and surrounding structures in dynamic supraclavicular ultrasound, supporting its potential clinical application in suprascapular neuropathy.
Automated ultrasound image classification is increasingly important for clinical decision support in breast, thyroid and fetal screening. However, deploying deep learning models in such safety-critical settings demands not only high predictive accuracy but also transparency, interpretability and trustworthiness-properties that existing approaches address insufficiently. Convolutional neural networks (CNNs) capture local texture patterns but struggle with global contextual dependencies, while Transformer-based models offer long-range reasoning yet require large-scale training data and remain sensitive to ultrasound-specific noise, both limiting factors for clinical deployment. We propose Multi-Scale CNN Token Transformer (MSCT-Trans), a lightweight and interpretable hybrid architecture for general-purpose ultrasound image classification. MSCT-Trans extracts multi-scale feature maps from a pre-trained CNN backbone and converts them into a unified token sequence, enabling a Transformer encoder to model global dependencies and inter-scale interactions over semantically meaningful, noise-attenuated representations. To support clinical transparency, we conducted a two-part explainability analysis-Gradient-weighted Class Activation Mapping++ spatial localisation and softmax class probability breakdown-demonstrating that MSCT-Trans consistently attends to diagnostically relevant anatomical regions, produces well-calibrated confidence estimates and associates prediction errors with model uncertainty rather than over-confident mis-classification. Here we evaluated MSCT-Trans on three ultrasound benchmarks spanning breast (BUS-BRA + BUSI + UCLM), thyroid (TN5000) and fetal imaging. MSCT-Trans consistently outperformed CNN and Transformer baselines across accuracy, macro-F1 and area under the receiver operating characteristic curve, particularly under class imbalance and limited data regimens. The combination of strong predictive performance, spatially grounded interpretability and calibrated uncertainty estimation positions MSCT-Trans as a transparent and trustworthy foundation for ultrasound-based clinical decision support. Code: https://github.com/MohsinFurkh/MSCT-Trans.
OBJECTIVE:This exploratory study aimed to develop a practical prediction model using clinical and sonographic features to identify patients with biopsy-proven ductal carcinoma in situ (DCIS) who are at low risk of pathological upstaging, with the goal of informing patient selection for DCIS active surveillance trial enrollment. METHODS:We retrospectively analyzed patients with DCIS diagnosed by core needle biopsy who underwent surgery at the National Cancer Center between February 2019 and December 2024. Clinical data and sonographic features were collected, along with selected mammographic and MRI variables for exploratory analysis. A predictive model was constructed by using multivariable logistic regression. RESULTS:We identified 224 patients diagnosed with DCIS through biopsy, including 96 pure DCIS cases (42.9%) and 128 DCIS cases with microinvasion (28.1%) or invasive carcinoma (29.0%) on final pathology. Multivariate analysis identified sonographic size (odds ratio [OR] 2.363, p = 0.02), palpable mass (OR 2.675, p = 0.02), non-parallel growth orientation on ultrasound (OR 4.449, p < 0.001), vascularity (Adler grade II-III) (OR 2.357, p = 0.014) and sonographically detected axillary lymphadenopathy (OR 5.262, p = 0.002) as independent predictors of upstaging. The predictive model constructed from these five variables achieved an area under the curve of 0.784 (95% confidence interval: 0.723-0.845) with overall accuracy of 72.3%. CONCLUSION:The proposed model based on routine clinical and sonographic features provided reasonable discrimination for upstaging risk in patients with biopsy-proven DCIS. It may serve as a useful exploratory reference for refining patient selection in active surveillance trial design.
OBJECTIVE:This research aims to evaluate the utility of ultrasound, quantitative computed tomography (QCT) and their integration for assessing muscle function and physical performance in older adults at risk of sarcopenia. METHODS:A total of 142 elderly inpatients were recruited in this cross-sectional study and completed physical performance tests. Participants were categorized into three parts: Part 1 (n = 125) underwent ultrasound to measure rectus femoris (RF) cross-sectional area (CSA), vastus lateralis pennation angle (PA) and gastrocnemius PA and fascicle length (Lf); Part 2 (n = 80) received clinically indicated QCT to quantify muscle fat fraction at T11, T12, L3 and L4; Part 3 (n = 63) completed both. Diagnostic parameters valuable for reduced muscle function from Parts 1 and 2 were integrated in Part 3 using binary logistic regression to construct the cross-modal model, with diagnostic performance assessed by area under the curve (AUC). RESULTS:Ultrasound-derived parameters (CSA of RF, PA of medial gastrocnemius and Lf of lateral gastrocnemius) and QCT-measured muscle fat fraction were significantly correlated with physical performance. Age, CSA of RF and fat fraction of psoas major muscle at L4 were identified as independent predictors of decreased gait speed and impaired physical performance. The cross-modal models integrating these parameters demonstrated improved diagnostic performance, with AUCs of 0.912 for decreased gait speed and 0.892 for impaired physical performance. CONCLUSION:The ultrasound and CT-derived muscle assessment system offers an objective imaging tool for evaluating muscle function and physical performance in older adults, and may serve as a valuable tool for sarcopenia auxiliary diagnosis.
Objective Ultrasound compression-wave attenuation (CWA) and shear-wave attenuation (SWA) quantify distinct mechanisms of ultrasonic energy loss in tissues and may provide complementary information for liver disease characterization. This study compares their diagnostic performance across the spectrum of metabolic dysfunction-associated steatotic liver disease (MASLD) and steatohepatitis (MASH). Methods In this prospective, cross-sectional study, 55 adults (13 healthy volunteers, 42 biopsy-proven MASLD patients) underwent ultrasound between 2020 and 2023. CWA and SWA were computed head-to-head on same acquisitions of participants. Correlations between CWA and SWA with histopathology and MRI-proton density fat fraction (PDFF) were assessed. Multivariate and area under the curve (AUC) analyses were performed. Results Both CWA and SWA correlated strongly with MRI-PDFF (r = 0.71 and 0.77, p < 0.001). Correlations with histopathology were 0.66 (steatosis), 0.61 (inflammation), 0.49 (ballooning), and 0.46 (fibrosis) for CWA; and 0.74, 0.53, 0.40, and 0.19 for SWA. Multivariable regressions confirmed independent associations with steatosis (CWA R² = 0.54, p < 0.001; SWA R² = 0.57, p < 0.001). For steatosis grading, AUCs for CWA versus SWA were 0.96 vs. 0.91 (S0 vs. ≥S1), 0.85 versus 0.87 (≤S1 vs. ≥S2), and 0.75 versus 0.88 (≤S2 vs. S3; p < 0.05) and for MASH detection the highest AUCs for CWA versus SWA were 0.94 versus 0.88. Conclusion CWA and SWA capture complementary mechanisms of energy loss, CWA reflecting lipid-related scattering and absorption, SWA reflecting viscoelastic dissipation. Both correlate strongly with MRI-PDFF and histopathology, though they show distinct secondary associations. Their combined assessment within a single ultrasound examination improves noninvasive steatosis grading and may reduce reliance on biopsy or MRI.
OBJECTIVE:Ultrasound is central to lymph node assessment and guides targeted biopsy. However, conventional Doppler has limited sensitivity to slow microvascular flow. Super-resolution ultrasound imaging using erythrocytes (SURE) is a fast, contrast-free technique that previously demonstrated feasibility in normal axillary and inguinal lymph nodes, but not in malignant lymph nodes or in the anatomically challenging cervical region. This feasibility study evaluated whether SURE could visualize the microvascular architecture in malignant cervical lymph nodes in patients with head and neck cancer or lymphoma, with an exploratory qualitative and quantitative comparison with normal cervical lymph nodes. METHODS:This preliminary feasibility study included five malignant cervical lymph nodes and five sonographically normal cervical lymph nodes. Conventional B-mode and Doppler images were obtained for qualitative comparison, followed by SURE acquisitions reconstructed into intensity and velocity maps. Exploratory quantitative parameters included vascular density, vessel diameter and peak flow velocity. RESULTS:SURE depicted the microvasculature in malignant cervical lymph nodes in greater detail than Doppler. Normal lymph nodes showed organized hilar branching patterns, whereas malignant lymph nodes showed heterogeneous and disorganized vascular patterns with focal avascular regions. In this small exploratory cohort, vascular density appeared lower in malignant than in normal lymph nodes (p = 0.009), while no statistically significant differences were observed in vessel diameter or peak flow velocity. CONCLUSION:SURE was feasible for fast, contrast-free microvascular imaging of malignant cervical lymph nodes. The findings are exploratory, and further methodological optimization and larger validation studies, including clinically indeterminate lymph nodes, are required to establish clinical utility.
Neuromuscular diseases (NMD), comprising over 600 different conditions, severely impact nerve and/or muscle function and lead to significant morbidity. Ultrasound is a non-invasive tool that is gaining acceptance for diagnosing NMD. In clinical practice, muscle ultrasound can be evaluated quantitatively or visually using an ordinal four-point grading score (Heckmatt score). Its current application is limited by time investment in manual analysis and lack of result transferability to other centers. Here, we present a single-center multi-modal deep learning framework using intermediate data fusion that improved the speed and diagnostic performance of muscle ultrasound. Our approach used neural networks enriched with patient-specific data of body mass index and age to predict neuromuscular pathology with an area under the precision-recall curve of 0.87 on a test set of 320 patients (220 with NMD and 100 in whom the diagnosis was refuted). SHAP analysis showed that adding BMI and age did not affect the model's performance. By leveraging Heckmatt scores from ultrasound images of six key muscles, our model efficiently and effectively identified the presence or absence of a neuromuscular disease. This approach may enhance the clinical utility of ultrasound by facilitating a more efficient diagnostic process for neuromuscular pathology, helping to guide the subsequent workup toward a definitive diagnosis.
Accurate perfusion evaluation is crucial for clinical diagnosis and treatment, such as lesion characterization and risk stratification of carotid atherosclerotic plaques. Contrast-enhanced ultrasound (CEUS) examination, which enhances microvascular visualization with contrast agents, offers the advantages of being non-invasive, real time and free of ionizing radiation, making it widely applicable for perfusion evaluation. However, traditional CEUS analysis may have several limitations, in part due to operator dependence. A single CEUS examination generates massive dynamic cine-loop sequences, so manual frame-by-frame analysis is not only time consuming and labor intensive, but also prone to missing key information. It also struggles to obtain quantitative perfusion parameters and has the problem of inter-observer variability. Artificial intelligence (AI) can efficiently process high-dimensional CEUS data through automated data pre-processing, intelligent segmentation of regions of interest, standardized feature extraction and accurate decision support. AI improves the diagnostic efficiency and consistency of CEUS analysis, thus becoming an ideal tool to address these limitations. In this review, we summarize the mechanism, clinical value and limitations of CEUS examination for perfusion evaluation and introduce the core AI techniques for CEUS analysis and the technical workflow of AI-assisted CEUS. We then elaborate on the status of applying AI-assisted CEUS across multiple organ systems. Finally, we discuss the current challenges and future directions of this technology, aiming to promote its clinical application.
OBJECTIVE:Detection of synovitis is essential for assessing rheumatoid arthritis (RA) activity and changing the therapy. This study aims to explore the level of agreement and correlation between Disease Activity Score 28 calculated with C-reactive protein levels (DAS28-CRP) and contrast-enhanced ultrasound (CEUS) and Superb microvascular imaging (SMI) in the classification of disease severity index in patients with RA who did not respond to second-line biologic therapy. METHODS:SMI and CEUS were applied to 37 patients with active RA does not respond to second-line biologic therapy. We evaluate the radiocarpal joint of both wrists. Differences in positive synovial vascularity (SV) and its semi-quantitative scale were observed, and the correlations of SMI and CEUS results with DAS-28. To obtain robust estimates and accurate confidence intervals, we conducted all analyses using nonparametric bootstrap methods. RESULTS:The results indicate that CEUS method shows fair agreement with DAS 28 clinical method (Kappa = 0.38), 95% CI (0.172, 0.547), p = 0.006, while SMI has weaker agreement (Kappa = 0.13) 95% CI (0.027, 0.246), p = 0.039. The correlation between CEUS and SMI is strong (ρ = 0.82), CI 95% (0.705, 0.879), suggesting that the two imaging modalities tend to produce similar classifications, although CEUS showed a stronger association with DAS28-CRP. CONCLUSION:Use of CEUS to detect vessels in the synovium and visualization of local SV is the method that most correlates with disease severity in relation to DAS 28 in patients with synovial arthritis who do not respond to second-line biologic therapy compared with SMI.
RATIONALE AND OBJECTIVES:Accurate quantification of fibroglandular tissue volume (FGV) and the fibroglandular ratio which is the ratio of FGV to the total breast volume (TBV) (FGR = FGV/TBV) is critical for assessment of a patient's lifetime breast cancer risk and personalized screening strategies. Magnetic resonance imaging (MRI) is considered the research standard for volumetric breast composition analysis but is limited by cost, availability, and contraindications. This study evaluates the capability of quantitative transmission (QT) imaging technology to measure FGV and FGR and compares these results with MRI-derived metrics, establishing quantitative equivalence between the two modalities. MATERIALS AND METHODS:A retrospective analysis included 53 breasts from 29 women imaged using QT scan and MRI within 30 days as part of the ACCRUE Study (ClinicalTrials.gov NCT03052166). QT data were acquired with the QT Scanner 2000 Model A and reconstructed into 3D speed-of-sound maps and respective reflection maps via a full-wave inverse scattering algorithm. FGV and FGR were computed using a fuzzy c-means segmentation as applied to the quantitative tissue maps. MRI datasets were segmented using a validated template-based method. Statistical comparisons included Pearson correlation (r), intraclass correlation coefficient (ICC), and Bland-Altman analysis. RESULTS:QT and MRI measurements demonstrated strong correlations for both FGV (r = 0.893) and FGR (r = 0.915, p < 0.0001). Mean biases were -13.2 cm³ for FGV and +12.7% for FGR. Regression slopes were 0.63 and 1.35, respectively, with significant proportional bias for both (p < 0.0001). Systematic differences were consistent with modality-specific contrast mechanisms and field-of-view definitions; the strong correlation and stable, predictable bias support quantitative concordance between the modalities. CONCLUSION:QT technology provides accurate, reproducible, and volumetric quantification of breast fibroglandular tissue, demonstrating strong concordance with MRI. These results supported FDA clearance of the QT FGR feature, establishing it as a validated quantitative imaging biomarker for clinical breast density assessment.
OBJECTIVE:To evaluate the puncture accuracy and feasibility of TONGMAI, an electromagnetic-tracking-based, ultrasound (US)-led multimodality imaging-guided robotic system using consistency-aware respiratory-guided cognitive fusion, in phantom and animal models. METHODS:Three experienced and six inexperienced physicians performed needle insertion in static phantom, respiratory phantom, New Zealand white rabbit liver and Guangxi Bama miniature-pig liver experiments using robot assistance, freehand puncture or needle-guide assistance. The primary outcome was real-time US-based puncture accuracy (US-Error). Exploratory robot-assisted punctures of non-hepatic targets (spleen, kidney, pancreas, gallbladder, bladder and lung) in miniature pigs were evaluated for first-attempt success, needle redirection, US-Error when applicable and complications. RESULTS:Among experienced physicians, US-Error did not differ significantly among robot assistance, freehand puncture and needle-guide assistance in static phantom, respiratory phantom, rabbit liver or miniature-pig liver experiments (overall p = 0.115, 0.131, 0.124 and 0.397, respectively). With robot assistance, mean US-Error values were 0.35, 0.57, 1.28 and 1.32 mm, respectively. Among inexperienced physicians, robot assistance reduced US-Error versus freehand puncture or needle-guide assistance in both phantom settings and versus freehand puncture in rabbit and miniature-pig liver experiments (p = 0.003 and p = 0.002, respectively). Under robot assistance, US-Error did not differ significantly between experienced and inexperienced physicians across all settings (p = 0.110, 0.054, 0.818 and 0.548, respectively). In exploratory non-hepatic punctures, first-attempt success varied by target, with no procedure-related complications. CONCLUSION:TONGMAI supported consistent puncture performance in experienced physicians and improved accuracy in inexperienced physicians under phantom and animal conditions, supporting further development of a US-led multimodality robotic workflow for deformable thoracoabdominal targets.
