Background The 2023 update of the American Association for the Study of Liver Diseases (AASLD version 2023 [hereafter, v2023]) guidance introduced new triggers for hepatocellular carcinoma (HCC) surveillance, incorporating US visualization score, increasing α-fetoprotein (AFP) level, and lesion growth, but has not yet been validated. Purpose To assess the surveillance performance of AASLD v2023 for HCC detection in comparison to Liver Imaging Reporting and Data System (LI-RADS) version 2017 (hereafter, v2017) and AASLD version 2018 (hereafter, v2018) in participants at high risk of HCC. Materials and Methods This prospective study consecutively enrolled high-risk participants undergoing US and AFP surveillance across three institutions between July 2023 and October 2024. Surveillance performance of US LI-RADS v2017, AASLD v2018, and AASLD v2023 for HCC detection was compared through sensitivity, specificity, positive predictive value, and negative predictive value (NPV). Multivariable logistic regression helped identify predictors of false-negative and false-positive classifications. Results Among 953 participants (median age, 51 years; IQR, 45-58 years; 658 men), 5% (50 of 953) had HCC. The new surveillance triggers (visualization score VIS-C, increasing AFP level, and lesion growth) demonstrated sensitivities ranging from 8% (four of 50 participants) to 48% (24 of 50 participants), with high specificities of 94% (849 of 903 participants) to 99.4% (898 of 903 participants). As an integrated algorithm, AASLD v2023 achieved a sensitivity of 94% (47 of 50 participants) and NPV of 99.6% (758 of 761 participants) for HCC detection, surpassing US LI-RADS v2017 (sensitivity, 60% [30 of 50 participants], P < .001; NPV, 98% [814 of 834 participants], P = .001) and AASLD v2018 (sensitivity, 76% [38 of 50 participants], P = .02; NPV, 98.5% [805 of 817 participants], P = .04). Specificity was lower than that with US LI-RADS v2017 (84% [758 of 903 participants] vs 90% [814 of 903 participants], P < .001) and AASLD v2018 (84% [758 of 903 participants] vs 89% [805 of 903 participants], P = .001). For early HCC, sensitivity remained superior to that of other algorithms (P < .001, P = .04). At multivariable analysis, AFP level of less than 20 ng/mL (odds ratio, 11.76; P < .001) and absence of cirrhosis (odds ratio, 2.45; P = .03) were independently associated with false-positive findings for AASLD v2023. Conclusion AASLD v2023 improved the sensitivity and NPV for HCC surveillance, outperforming US LI-RADS v2017 and AASLD v2018. Chinese Clinical Trial Registry no. ChiCTR2100054330 © RSNA, 2026 Supplemental material is available for this article.

Liver fibrosis is a reversible precursor to cirrhosis, and early detection is key to halting disease progression. Tongue diagnosis provides a non-invasive and cost-effective insight into internal health; however, its subjectivity limits clinical reliability. We developed TongVMoe, a multi-task deep learning model trained on 2202 tongue images from 1601 patients, to detect liver fibrosis and simultaneously classify seven key tongue features. The model achieved an area under the curve (AUC) of 0.8061, outperforming State-of-the-Art methods such as DiffMIC-v2 (0.6929), HorNet (0.7018), InceptionNeXt (0.7012), LSNet (0.6971), and TransXNet (0.7062). TongVMoe also demonstrated robust recognition of tongue features, with AUCs of 0.9752 for cracks and 0.9232 for greasy coating. Among these features, petechiae emerged as a significant clinical indicator, showing a strong correlation with liver fibrosis (χ² = 19.516, P < 0.001). We further integrated the model into a WeChat mini-program and simulated remote screening, achieving an accuracy of 77.8% and a sensitivity of 86.2%. These findings suggest that the TongVMoe has the potential to serve as an interpretable and mobile-compatible tool for the early detection and monitoring of liver fibrosis, particularly in resource-limited areas. Trial registration: Chinese Clinical Trial Registry (ChiCTR2100053676, registered 27 November 2021).

PurposeThis study aims to develop an autonomous robotic ultrasound scanning system (auto-RUSS) pipeline, comparing its reproducibility and observer consistency in image analysis with physicians of varying levels of expertise.Design/methodology/approachAn auto-RUSS was engineered using a 7-degree-of-freedom robotic arm, with real-time regulation based on force control and ultrasound visual servoing. Two phantoms were employed for the human-machine comparative experiment, involving three groups: auto-RUSS, non-expert (4 junior physicians), and expert (4 senior physicians). This setup enabled comprehensive assessment of reproducibility in contact force, image acquisition, image measurement and AI-assisted classification. Radiological feature variability was measured using the coefficient of variation (COV), while performance and reproducibility assessments utilized mean and standard deviation (SD).FindingsThe auto-RUSS had the potential to reduce operator-dependent variability in ultrasound examinations, offering enhanced repeatability and consistency across multiple dimensions including probe contact force, images acquisition, image measurement, and diagnostic model performance.Originality/valueIn this paper, an autonomous robotic ultrasound scanning system (auto-RUSS) pipeline was proposed. Through comprehensive human-machine comparison experiments, the auto-RUSS was shown to effectively improve the reproducibility of ultrasound images and minimize human-induced variability.

Data heterogeneity critically limits distributed artificial intelligence (AI) in medical imaging. We propose HeteroSync Learning (HSL), a privacy-preserving framework that addresses heterogeneity through: (1) Shared Anchor Task (SAT) for cross-node representation alignment, and (2) an Auxiliary Learning Architecture coordinating SAT with local primary tasks. Validated via large-scale simulations (feature/label/quantity/combined heterogeneity) and a real-world multi-center thyroid cancer study, HSL outperforms local learning, 12 benchmark methods (FedAvg, FedProx, SplitAVG, FedRCL, FedCOME, etc.), and foundation models (e.g., CLIP) by better stability and up to 40% in area under the curve (AUC), matching central learning performance. HSL achieves 0.846 AUC on the out-of-distribution pediatric thyroid cancer data (outperforming others by 5.1-28.2%), demonstrating superior generalization. Visualizations confirm HSL successfully homogenizes heterogeneous distributions. This work provides an effective solution for distributed medical AI, enabling equitable collaboration across institutions and advancing healthcare AI democratization.

OBJECTIVES:This study aimed to investigate and compare 2-dimensional shear wave elastography (2D-SWE) measurements and influencing factors among 2 different devices and to evaluate the ability and influencing factors of these measurements to assess liver fibrosis. METHODS:From October 2022 to September 2023, 290 hepatocellular carcinoma (HCC) patients and 30 healthy volunteers were prospectively included. The 2D-SWE measurements were performed using AixPlorer V (SEmean) and APLIO i900 (CEmean). This study compared 2D-SWE measurements between instruments for evaluating the liver fibrosis stage and analyzed the potential influencing factors. RESULTS:The 2D-SWE measurements obtained by the 2 instruments were significantly different (P < .001), but the differences were significant only for patients with stage F4 liver fibrosis (P < .001) and not for volunteers or patients with stage F0-F3 liver fibrosis (all P > .050). Multivariate linear regression analysis revealed that the factors independently influencing the SEmean were alanine aminotransferase (ALT) (P = .034) and liver fibrosis stage (P < .001), while fibrosis stage (P = .028) was the only factor influencing the CEmean. CONCLUSIONS:Although 2D-SWE from the 2 different instruments was capable of detecting liver fibrosis, it yielded varying results in HCC patients. These discrepancies were predominantly observed in patients with F4 liver fibrosis but not in healthy adults or patients with F0-F3 liver fibrosis. One potential contributing factor to the differences between instruments could be ALT levels.

Background Large language models (LLMs) hold substantial promise for medical imaging interpretation. However, there is a lack of studies on their feasibility in handling reasoning questions associated with medical diagnosis. Purpose To investigate the viability of leveraging three publicly available LLMs to enhance consistency and diagnostic accuracy in medical imaging based on standardized reporting, with pathology as the reference standard. Materials and Methods US images of thyroid nodules with pathologic results were retrospectively collected from a tertiary referral hospital between July 2022 and December 2022 and used to evaluate malignancy diagnoses generated by three LLMs-OpenAI's ChatGPT 3.5, ChatGPT 4.0, and Google's Bard. Inter- and intra-LLM agreement of diagnosis were evaluated. Then, diagnostic performance, including accuracy, sensitivity, specificity, and area under the receiver operating characteristic curve (AUC), was evaluated and compared for the LLMs and three interactive approaches: human reader combined with LLMs, image-to-text model combined with LLMs, and an end-to-end convolutional neural network model. Results A total of 1161 US images of thyroid nodules (498 benign, 663 malignant) from 725 patients (mean age, 42.2 years ± 14.1 [SD]; 516 women) were evaluated. ChatGPT 4.0 and Bard displayed substantial to almost perfect intra-LLM agreement (κ range, 0.65-0.86 [95% CI: 0.64, 0.86]), while ChatGPT 3.5 showed fair to substantial agreement (κ range, 0.36-0.68 [95% CI: 0.36, 0.68]). ChatGPT 4.0 had an accuracy of 78%-86% (95% CI: 76%, 88%) and sensitivity of 86%-95% (95% CI: 83%, 96%), compared with 74%-86% (95% CI: 71%, 88%) and 74%-91% (95% CI: 71%, 93%), respectively, for Bard. Moreover, with ChatGPT 4.0, the image-to-text-LLM strategy exhibited an AUC (0.83 [95% CI: 0.80, 0.85]) and accuracy (84% [95% CI: 82%, 86%]) comparable to those of the human-LLM interaction strategy with two senior readers and one junior reader and exceeding those of the human-LLM interaction strategy with one junior reader. Conclusion LLMs, particularly integrated with image-to-text approaches, show potential in enhancing diagnostic medical imaging. ChatGPT 4.0 was optimal for consistency and diagnostic accuracy when compared with Bard and ChatGPT 3.5. © RSNA, 2024 Supplemental material is available for this article.

Background Tongue inspection, an essential diagnostic method in Traditional Chinese Medicine (TCM), has the potential for early-stage disease screening. This study aimed to evaluate the effectiveness of deep learning-based analysis of tongue images for hepatic fibrosis screening. Methods A total of 1083 tongue images were collected from 741 patients and divided into training, validation, and test sets. DenseNet-201, a convolutional neural network, was employed to train the AI model using these tongue images. The predictive performance of AI was assessed and compared with that of FIB-4, using real-time two-dimensional shear wave elastography as the reference standard. Results The proposed AI model achieved an accuracy of 0.845 (95% CI: 0.79–0.90) and 0.814 (95% CI: 0.76–0.87) in the validation and test sets, respectively, with negative predictive values (NPVs) exceeding 90% in both sets. The AI model outperformed FIB-4 in all aspects, and when combined with FIB-4, the NPV reached 94.4%. Conclusion Tongue inspection, with the assistance of AI, could serve as a first-line screening method for hepatic fibrosis.

To establish a nomogram for differentiating malignant and benign focal liver lesions (FLLs) using ultrasomics features derived from contrast-enhanced ultrasound (CEUS). 527 patients were retrospectively enrolled. On the training cohort, ultrasomics features were extracted from CEUS and b-mode ultrasound (BUS). Automatic feature selection and model development were performed using the Ultrasomics-Platform software, outputting the corresponding ultrasomics scores. A nomogram based on the ultrasomics scores from artery phase (AP), portal venous phase (PVP) and delayed phase (DP) of CEUS, and clinical factors were established. On the validation cohort, the diagnostic performance of the nomogram was assessed and compared with seniorexpert and resident radiologists. In the training cohort, the AP, PVP and DP scores exhibited better differential performance than BUS score, with area under the curve (AUC) of 84.1-85.1

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.

Background: Noninvasive tests can be used to screen patients with chronic liver disease for advanced liver fibrosis; however, the use of single tests may not be adequate. Purpose: To construct sequential clinical algorithms that include a US deep learning (DL) model and compare their ability to predict advanced liver fibrosis with that of other noninvasive tests. Materials and Methods: This retrospective study included adult patients with a history of chronic liver disease or unexplained abnormal liver function test results who underwent B -mode US of the liver between January 2014 and September 2022 at three health care facilities. A US -based DL network (FIB -Net) was trained on US images to predict whether the shear -wave elastography (SWE) value was 8.7 kPa or higher, indicative of advanced fibrosis. In the internal and external test sets, a two-step algorithm (Two-step#1) using the Fibrosis -4 Index (FIB -4) followed by FIB -Net and a three -step algorithm (Three-step#1) using FIB -4 followed by FIB -Net and SWE were used to simulate screening scenarios where liver stiffness measurements were not or were available, respectively. Measures of diagnostic accuracy were calculated using liver biopsy as the reference standard and compared between FIB -4, SWE, FIB -Net, and European Association for the Study of the Liver guidelines (ie, FIB -4 followed by SWE), along with sequential algorithms. Results: The training, validation, and test data sets included 3067 (median age, 42 years [IQR, 33-53 years]; 2083 male), 1599 (median age, 41 years [IQR, 33-51 years]; 1124 male), and 1228 (median age, 44 years [IQR, 33-55 years]; 741 male) patients, respectively. FIB -Net obtained a noninferior specificity with a margin of 5% ( P < .001) compared with SWE (80% vs 82%). The Two-step#1 algorithm showed higher specificity and positive predictive value (PPV) than FIB -4 (specificity, 79% vs 57%; PPV, 44% vs 32%) while reducing unnecessary referrals by 42%. The Three-step#1 algorithm had higher specificity and PPV compared with European Association for the Study of the Liver guidelines (specificity, 94% vs 88%; PPV, 73% vs 64%) while reducing unnecessary referrals by 35%. Conclusion: A sequential algorithm combining FIB -4 and a US DL model showed higher diagnostic accuracy and improved referral management for all -cause advanced liver fibrosis compared with FIB -4 or the DL model alone.

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.

Importance:To optimize the integration of artificial intelligence (AI) decision aids and reduce workload in thyroid nodule management, it is critical to incorporate personalized AI into the decision-making processes of radiologists with varying levels of expertise.Objective:To develop an optimized integration of AI decision aids for reducing radiologists' workload while maintaining diagnostic performance compared with traditional AI-assisted strategy.Design, Setting, and Participants:In this diagnostic study, a retrospective set of 1754 ultrasonographic images of 1048 patients with 1754 thyroid nodules from July 1, 2018, to July 31, 2019, was used to build an optimized strategy based on how 16 junior and senior radiologists incorporated AI-assisted diagnosis results with different image features. In the prospective set of this diagnostic study, 300 ultrasonographic images of 268 patients with 300 thyroid nodules from May 1 to December 31, 2021, were used to compare the optimized strategy with the traditional all-AI strategy in terms of diagnostic performance and workload reduction. Data analyses were completed in September 2022.Main Outcomes and Measures:The retrospective set of images was used to develop an optimized integration of AI decision aids for junior and senior radiologists based on the selection of AI-assisted significant or nonsignificant features. In the prospective set of images, the diagnostic performance, time-based cost, and assisted diagnosis were compared between the optimized strategy and the traditional all-AI strategy.Results:The retrospective set included 1754 ultrasonographic images from 1048 patients (mean [SD] age, 42.1 [13.2] years; 749 women [71.5%]) with 1754 thyroid nodules (mean [SD] size, 16.4 [10.6] mm); 748 nodules (42.6%) were benign, and 1006 (57.4%) were malignant. The prospective set included 300 ultrasonographic images from 268 patients (mean [SD] age, 41.7 [14.1] years; 194 women [72.4%]) with 300 thyroid nodules (mean [SD] size, 17.2 [6.8] mm); 125 nodules (41.7%) were benign, and 175 (58.3%) were malignant. For junior radiologists, the ultrasonographic features that were not improved by AI assistance included cystic or almost completely cystic nodules, anechoic nodules, spongiform nodules, and nodules smaller than 5 mm, whereas for senior radiologists the features that were not improved by AI assistance were cystic or almost completely cystic nodules, anechoic nodules, spongiform nodules, very hypoechoic nodules, nodules taller than wide, lobulated or irregular nodules, and extrathyroidal extension. Compared with the traditional all-AI strategy, the optimized strategy was associated with increased mean task completion times for junior radiologists (reader 11, from 15.2 seconds [95% CI, 13.2-17.2 seconds] to 19.4 seconds [95% CI, 15.6-23.3 seconds]; reader 12, from 12.7 seconds [95% CI, 11.4-13.9 seconds] to 15.6 seconds [95% CI, 13.6-17.7 seconds]), but shorter times for senior radiologists (reader 14, from 19.4 seconds [95% CI, 18.1-20.7 seconds] to 16.8 seconds [95% CI, 15.3-18.3 seconds]; reader 16, from 12.5 seconds [95% CI, 12.1-12.9 seconds] to 10.0 seconds [95% CI, 9.5-10.5 seconds]). There was no significant difference in sensitivity (range, 91%-100%) or specificity (range, 94%-98%) between the 2 strategies for readers 11 to 16.Conclusions and Relevance:This diagnostic study suggests that an optimized AI strategy in thyroid nodule management may reduce diagnostic time-based costs without sacrificing diagnostic accuracy for senior radiologists, while the traditional all-AI strategy may still be more beneficial for junior radiologists.

objective

Purpose:The T cell-inflamed gene expression profile (GEP) quantifies 18 genes' expression indicative of a T-cell immune tumor microenvironment, playing a crucial role in the immunotherapy of hepatocellular carcinoma (HCC). Our study aims to develop a radiomics-based machine learning model using contrast-enhanced ultrasound (CEUS) for predicting T cell-inflamed GEP in HCC.Methods:The primary cohort of HCC patients with preoperative CEUS and RNA sequencing data of tumor tissues at the single center was used to construct the model. A total of 5936 radiomics features were extracted from the regions of interest in representative images of each phase, and the least absolute shrinkage and selection operator and logistic regression were used to construct four models including three phase-specific models and an integrated model. The area under the curve (AUC) was calculated to evaluate the performance of the model. The independent cohort of HCC patients with preoperative CEUS and Immunoscore based on immunohistochemistry and digital pathology was used to validate the correlation between model prediction value and T-cell infiltration.Results:There were 268 patients enrolled in the primary cohort and 46 patients enrolled in the independent cohort. Compared with the other three models, the AP model constructed by 36 arterial phase (AP) features showed good performance with a mean AUC of 0.905 in the 5-fold cross-validation and was easier to apply in the clinical setting. The decision curve and calibration curve confirmed the clinical utility of the model. In the independent cohort, patients with high Immunoscores showed significantly higher GEP prediction values than those with low Immunoscores (t=-2.359, p=0.029).Conclusion:The CEUS-based model is a reliable predictive tool for T cell-inflamed GEP in HCC, and might facilitate individualized immunotherapy decision-making.

To develop and validate a nomogram incorporating circulating tumour cell counts (CTCs) and the ultrasomics signatures of contrast-enhancement ultrasound (CEUS) for predicting postoperative early recurrence (ER) of HCC after radical treatment.Methods Between December 2017 and December 2018, 153 HCC patients (134 males and 19 females; mean age, 56.0 +/- 10.2 years; range, 28-78 years) treated with radical therapy were enrolled in our retrospective study and were divided into a training cohort (n = 107) and a validation cohort (n = 46). All patients underwent preoperative CTC tests and CEUS examinations before treatment. The ultrasomics signature was extracted and built from CEUS images. Univariate and multivariate logistic regression analyses were used to identify the significant variables related to ER, which were then combined to build a predictive nomogram. The performance of the nomogram was evaluated by its discrimination, calibration and clinical utility. The predictive model was further evaluated in the internal validation cohort.Results HBV DNA, serum AFP level, CTC status, tumour size and ultrasomics score were identified as independent predictors associated with ER (all p < 0.05). Multivariable logistic regression analysis showed that the CTC status (OR = 7.02 [95% CI, 2.07 to 28.38], p = 0.003) and ultrasomics score (OR = 148.65 [95% CI, 25.49 to 1741.72], p < 0.001) were independent risk factors for ER. The nomogram based on ultrasomics score, CTC status, serum AFP level and tumour size exhibited C-indexes of 0.933 (95% CI, 0.878 to 0.988) and 0.910 (95% CI, 0.765 to 1.055) in the training and validation cohorts, respectively, fitting well in calibration curves. Decision curve analysis further confirmed the clinical usefulness of the nomogram.Conclusion The nomogram incorporating CTC, ultrasomics features and independent clinical risk factors achieved satisfactory preoperative prediction of ER in HCC patients after radical treatment.Advances in knowledge 1. CTC status and ultrasomics score were identified as independent predictors associated with ER of HCC after radical treatment. 2. The nomogram constructed by ultrasomics score generated by 17 ultrasomics features, combined with CTCs and independent clinical risk factors such as AFP and tumour size. 3. The nomogram exhibited satisfactory discriminative power, and could be clinically useful in the preoperative prediction of ER after radical treatment in HCC patients.

Contrast-enhanced ultrasound (CEUS) is generally employed for focal liver lesions (FLLs) diagnosis. Among the FLLs, atypical hepatocellular carcinoma (HCC) is difficult to distinguish from focal nodular hyperplasia (FNH) in CEUS video. For this reason, we propose and evaluate a feature fusion method to resolve this problem. The proposed algorithm extracts a set of hand-crafted features and the deep features from the CEUS cine clip data. The hand-crafted features include the spatial-temporal feature based on a novel descriptor called Velocity-Similarity and Dissimilarity Matching Local Binary Pattern (V-SDMLBP), and the deep features from a 3-D convolution neural network (3D-CNN). Then the two types of features are fused. Finally, a classifier is employed to diagnose HCC or FNH. Several classifiers have achieved excellent performance, which demonstrates the superiority of the fused features. In addition, compared with general CNNs, the proposed fused features have better interpretability.

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, which are often overlooked by existing deep learning methods. In this study, we propose 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. First, we propose a hybrid-fusion based multimodal framework that integrates B-mode, SWE, CEUS ultrasound data and clinical data simultaneously for the first time for the diagnosis of hepatocellular carcinoma and intrahepatic cholangiocarcinoma. Furthermore, we design "Lesions Pairing" that can construct the data set to solve the problem of small samples in multi-modal learning. The experimental results demonstrate that the area under the curve of the hybrid-fusion framework is 0.9854, which is better than that of other single-modalities and fusion strategies.

To improve noninvasive diagnosis of HCC using a combination of CE US LI-RADS and alpha-fetoprotein (AFP). 757 solitary liver nodules from 757 patients at risk of HCC with CE US and serum AFP test were categorized as LR-1 to LR-5 through LR-M according to CE US LI-RADS version 2017. In LR-3, LR-4, and LR-M nodules, those with AFP > 200 ng/ml were reclassified as mLR-5. Nodules with LR-5 and mLR-5 were reclassified as definitely HCC to modify CE US LI-RADS. Diagnostic performance was assessed with specificity, sensitivity, and PPV. The sensitivity, specificity, and PPV of LR-5 as a predictor of HCC were 64.7%, 97.8%, and 98.9%, respectively. 32.1% patients with solitary liver nodule had AFP greater than 200 ng/ml, of which 98.8% were HCC (25.8%, 7.5%, 2.5% assigned to LR-M, LR-4, LR-3, respectively) and 1.2% were Combined Hepatocellular Cholangiocarcinoma. After modification, the sensitivity increased to 79.6% (P < 0.001), while specificity and PPV remained high (96.6% and 98.7%, P > 0.050). The combination of CE US LI-RADS and AFP for diagnosing HCC improved diagnostic sensitivity significantly, while maintaining high PPV and specificity in patients with the solitary liver nodule.