ObjectiveDiscriminating atypical hepatocellular carcinoma (HCC) from other malignancies in liver nodules classified as Liver Imaging Reporting and Data System category M (LR-M) remains a significant diagnostic challenge on conventional ultrasound examination. The LR-M category, originally intended to capture non-HCC malignancies, paradoxically contains up to 63% of atypical HCCs that deviate from classic enhancement patterns, leading to potential misdiagnosis and suboptimal treatment planning. While deep learning has shown promise in HCC diagnosis, most existing models rely exclusively on single-modality ultrasound, overlooking the diagnostic benefits of integrating complementary information from multiple imaging sources. To address this gap, we propose a novel attention-weighted tri-modal ultrasound network (TUS-Net) that integrates contrast-enhanced ultrasound (CEUS), B-mode ultrasound (BUS), and time-intensity curves (TICs) to improve diagnostic accuracy for these clinically challenging lesions. MethodsOur framework incorporates a three-dimensional convolutional neural network (C3D) backbone to extract spatiotemporal features from CEUS videos, capturing dynamic vascular patterns critical for lesion characterization. To effectively fuse complementary modalities, we introduce a dual-channel feature fusion module (DCFFM) that adaptively combines features from CEUS and BUS through channel-wise attention mechanisms, allowing the model to dynamically weigh the contribution of each modality based on diagnostic relevance. Additionally, we propose a temporal intensity feature fusion module (TIFFM) that leverages quantitative hemodynamic information from TICs to guide the model’s attention toward diagnostically critical temporal phases, such as arterial wash-in and portal venous washout. The model is further enhanced by automated lesion localization using YOLOX and class activation mapping for interpretability, ensuring that predictions align with clinically meaningful imaging features. ResultsEvaluated on a tri-modal ultrasound dataset comprising 161 patients with pathologically confirmed LR-M nodules (131 atypical HCC and 30 non-HCC malignancies), our model achieved an accuracy of 86.83%, a sensitivity of 92.50%, a specificity of 75.50%, and an AUC of 89.32% in screening atypical HCC. Compared to single-modality baselines, TUS-Net demonstrated superior specificity, a clinically critical metric given the higher risk associated with misclassifying non-HCC malignancies. Ablation studies confirmed the contribution of each module, with the full model outperforming both standard C3D and 3D ResNet backbones integrated with attention mechanisms. A reader study involving junior and senior radiologists further validated the clinical utility of AI assistance, showing consistent improvements in specificity and inter-reader consistency, particularly for less experienced clinicians. ConclusionThese results surpass existing benchmark models and demonstrate the potential of our approach to enhance diagnostic precision in clinically specific cases. By intelligently fusing multi-modal ultrasound data with attention-guided mechanisms, TUS-Net offers a reliable and interpretable tool that holds promise for improving the non-invasive diagnosis of atypical HCC in challenging LR-M liver nodules.

ObjectiveDiscriminating atypical hepatocellular carcinoma (HCC) from other malignancies in liver nodules classified as Liver Imaging Reporting and Data System category M (LR-M) remains a significant diagnostic challenge on conventional ultrasound examination. The LR-M category, originally intended to capture non-HCC malignancies, paradoxically contains up to 63% of atypical HCCs that deviate from classic enhancement patterns, leading to potential misdiagnosis and suboptimal treatment planning. While deep learning has shown promise in HCC diagnosis, most existing models rely exclusively on single-modality ultrasound, overlooking the diagnostic benefits of integrating complementary information from multiple imaging sources. To address this gap, we propose a novel attention-weighted tri-modal ultrasound network (TUS-Net) that integrates contrast-enhanced ultrasound (CEUS), B-mode ultrasound (BUS), and time-intensity curves (TICs) to improve diagnostic accuracy for these clinically challenging lesions. MethodsOur framework incorporates a three-dimensional convolutional neural network (C3D) backbone to extract spatiotemporal features from CEUS videos, capturing dynamic vascular patterns critical for lesion characterization. To effectively fuse complementary modalities, we introduce a dual-channel feature fusion module (DCFFM) that adaptively combines features from CEUS and BUS through channel-wise attention mechanisms, allowing the model to dynamically weigh the contribution of each modality based on diagnostic relevance. Additionally, we propose a temporal intensity feature fusion module (TIFFM) that leverages quantitative hemodynamic information from TICs to guide the model’s attention toward diagnostically critical temporal phases, such as arterial wash-in and portal venous washout. The model is further enhanced by automated lesion localization using YOLOX and class activation mapping for interpretability, ensuring that predictions align with clinically meaningful imaging features. ResultsEvaluated on a tri-modal ultrasound dataset comprising 161 patients with pathologically confirmed LR-M nodules (131 atypical HCC and 30 non-HCC malignancies), our model achieved an accuracy of 86.83%, a sensitivity of 92.50%, a specificity of 75.50%, and an AUC of 89.32% in screening atypical HCC. Compared to single-modality baselines, TUS-Net demonstrated superior specificity, a clinically critical metric given the higher risk associated with misclassifying non-HCC malignancies. Ablation studies confirmed the contribution of each module, with the full model outperforming both standard C3D and 3D ResNet backbones integrated with attention mechanisms. A reader study involving junior and senior radiologists further validated the clinical utility of AI assistance, showing consistent improvements in specificity and inter-reader consistency, particularly for less experienced clinicians. ConclusionThese results surpass existing benchmark models and demonstrate the potential of our approach to enhance diagnostic precision in clinically specific cases. By intelligently fusing multi-modal ultrasound data with attention-guided mechanisms, TUS-Net offers a reliable and interpretable tool that holds promise for improving the non-invasive diagnosis of atypical HCC in challenging LR-M liver nodules.

Progressive pulmonary fibrosis (PPF) is a progressive and lethal condition with few effective treatment options. Improvements in quality of life for patients with PPF remain limited even while receiving treatment with approved antifibrotic drugs. Traditional Chinese medicine (TCM) has the potential to improve cough, dyspnea and fatigue symptoms of patients with PPF. TCM treatments are typically diverse and individualized, requiring urgent development of efficient and precise design strategies to identify effective treatment options. We designed an innovative Bayesian adaptive two-stage trial, hoping to provide new ideas for the rapid evaluation of the effectiveness of TCM in PPF. An open-label, two-stage, adaptive Bayesian randomized controlled trial will be conducted in China. Based on Bayesian methods, the trial will employ response-adaptive randomization to allocate patients to study groups based on data collected over the course of the trial. The adaptive Bayesian trial design will employ a Bayesian hierarchical model with "stopping" and "continuation" criteria once a predetermined posterior probability of superiority or futility and a decision threshold are reached. The trial can be implemented more efficiently by sharing the master protocol and organizational management mechanisms of the sub-trial we have implemented. The primary patient-reported outcome is a change in the Leicester Cough Questionnaire score, reflecting an improvement in cough-specific quality of life. The adaptive Bayesian trial design may be a promising method to facilitate the rapid clinical evaluation of TCM effectiveness for PPF, and will provide an example for how to evaluate TCM effectiveness in rare and refractory diseases. However, due to the complexity of the trial implementation, sufficient simulation analysis by professional statistical analysts is required to construct a Bayesian response-adaptive randomization procedure for timely response. Moreover, detailed standard operating procedures need to be developed to ensure the feasibility of the trial implementation. Please cite this article as: Zhang C, Nie YS, Zhang CT, Yang HJ, Zhang HR, Xiao W, Cui GF, Li J, Li SJ, Huang QS, Yan SY. An adaptive Bayesian randomized controlled trial of traditional Chinese medicine in progressive pulmonary fibrosis: Rationale and study design. J Integr Med. 2025; 23(2): 138-145.
Female ; Humans ; Male ; Bayes Theorem ; Disease Progression ; Drugs, Chinese Herbal/therapeutic use* ; Medicine, Chinese Traditional/methods* ; Pulmonary Fibrosis/therapy* ; Quality of Life ; Randomized Controlled Trials as Topic ; Research Design ; Adaptive Clinical Trials as Topic

OBJECTIVE: To assess the independent and combined effects of air pollutants, meteorological factors, and greenspace exposure on new tuberculosis (TB) cases. METHODS: TB case data from Shanghai (2013-2018) were obtained from the Shanghai Center for Disease Control and Prevention. Environmental data on air pollutants, meteorological variables, and greenspace exposure were obtained from the National Tibetan Plateau Data Center. We employed a distributed-lag nonlinear model to assess the effects of these environmental factors on TB cases. RESULTS: Increased TB risk was linked to PM _2.5, PM ₁₀, and rainfall, whereas NO ₂, SO ₂, and air pressure were associated with a reduced risk. Specifically, the strongest cumulative effects occurred at various lags: PM _2.5 ( RR = 1.166, 95% CI: 1.026-1.325) at 0-19 weeks; PM ₁₀ ( RR = 1.167, 95% CI: 1.028-1.324) at 0-18 weeks; NO ₂ ( RR = 0.968, 95% CI: 0.938-0.999) at 0-1 weeks; SO ₂ ( RR = 0.945, 95% CI: 0.894-0.999) at 0-2 weeks; air pressure ( RR = 0.604, 95% CI: 0.447-0.816) at 0-8 weeks; and rainfall ( RR = 1.404, 95% CI: 1.076-1.833) at 0-22 weeks. Green space exposure did not significantly impact TB cases. Additionally, low temperatures amplified the effect of PM _2.5 on TB. CONCLUSION Exposure to PM _2.5, PM ₁₀, and rainfall increased the risk of TB, highlighting the need to address air pollutants for the prevention of TB in Shanghai.
China/epidemiology* ; Humans ; Air Pollutants/analysis* ; Tuberculosis/epidemiology* ; Incidence ; Meteorological Concepts ; Particulate Matter/adverse effects* ; Environmental Exposure ; Male ; Female ; Adult ; Air Pollution ; Middle Aged

Ectodermal dysplasia is a group of hereditary diseases characterized by developmental defects of ectodermal structures. Its oral manifestations mainly center on congenital missing teeth, abnormal tooth morphology, and maxillofacial bone developmental disorders, which seriously affect the masticatory function, maxillofacial development, and mental health of affected children. In this article, the multidimensional diagnostic strategy system for children with ectodermal dysplasia and the related progress of early oral prosthodontic treatment methods were systematically reviewed to provide references for clinicians in the diagnosis and treatment of children with ectodermal dysplasia.
Child ; Humans ; Anodontia ; Ectodermal Dysplasia/diagnosis* ; Prosthodontics ; Tooth Abnormalities/therapy*

ObjectivePrimary liver cancer, predominantly hepatocellular carcinoma (HCC), is a significant global health issue, ranking as the sixth most diagnosed cancer and the third leading cause of cancer-related mortality. Accurate and early diagnosis of HCC is crucial for effective treatment, as HCC and non-HCC malignancies like intrahepatic cholangiocarcinoma (ICC) exhibit different prognoses and treatment responses. Traditional diagnostic methods, including liver biopsy and contrast-enhanced ultrasound (CEUS), face limitations in applicability and objectivity. The primary objective of this study was to develop an advanced, light-weighted classification network capable of distinguishing HCC from other non-HCC malignancies by leveraging the automatic analysis of brightness changes in CEUS images. The ultimate goal was to create a user-friendly and cost-efficient computer-aided diagnostic tool that could assist radiologists in making more accurate and efficient clinical decisions. MethodsThis retrospective study encompassed a total of 161 patients, comprising 131 diagnosed with HCC and 30 with non-HCC malignancies. To achieve accurate tumor detection, the YOLOX network was employed to identify the region of interest (ROI) on both B-mode ultrasound and CEUS images. A custom-developed algorithm was then utilized to extract brightness change curves from the tumor and adjacent liver parenchyma regions within the CEUS images. These curves provided critical data for the subsequent analysis and classification process. To analyze the extracted brightness change curves and classify the malignancies, we developed and compared several models. These included one-dimensional convolutional neural networks (1D-ResNet, 1D-ConvNeXt, and 1D-CNN), as well as traditional machine-learning methods such as support vector machine (SVM), ensemble learning (EL), k-nearest neighbor (KNN), and decision tree (DT). The diagnostic performance of each method in distinguishing HCC from non-HCC malignancies was rigorously evaluated using four key metrics: area under the receiver operating characteristic (AUC), accuracy (ACC), sensitivity (SE), and specificity (SP). ResultsThe evaluation of the machine-learning methods revealed AUC values of 0.70 for SVM, 0.56 for ensemble learning, 0.63 for KNN, and 0.72 for the decision tree. These results indicated moderate to fair performance in classifying the malignancies based on the brightness change curves. In contrast, the deep learning models demonstrated significantly higher AUCs, with 1D-ResNet achieving an AUC of 0.72, 1D-ConvNeXt reaching 0.82, and 1D-CNN obtaining the highest AUC of 0.84. Moreover, under the five-fold cross-validation scheme, the 1D-CNN model outperformed other models in both accuracy and specificity. Specifically, it achieved accuracy improvements of 3.8% to 10.0% and specificity enhancements of 6.6% to 43.3% over competing approaches. The superior performance of the 1D-CNN model highlighted its potential as a powerful tool for accurate classification. ConclusionThe 1D-CNN model proved to be the most effective in differentiating HCC from non-HCC malignancies, surpassing both traditional machine-learning methods and other deep learning models. This study successfully developed a user-friendly and cost-efficient computer-aided diagnostic solution that would significantly enhances radiologists’ diagnostic capabilities. By improving the accuracy and efficiency of clinical decision-making, this tool has the potential to positively impact patient care and outcomes. Future work may focus on further refining the model and exploring its integration with multimodal ultrasound data to maximize its accuracy and applicability.

Objective To investigate changes in the urinary metabolite profiles of children exposed to polycyclic aromatic hydrocarbons(PAHs)during critical brain development and explore their potential link with the intestinal microbiota. Methods Liquid chromatography-tandem mass spectrometry was used to determine ten hydroxyl metabolites of PAHs(OH-PAHs)in 36-month-old children.Subsequently,37 children were categorized into low-and high-exposure groups based on the sum of the ten OH-PAHs.Ultra-high-performance liquid chromatography coupled with quadrupole time-of-flight mass spectrometry was used to identify non-targeted metabolites in the urine samples.Furthermore,fecal flora abundance was assessed by 16S rRNA gene sequencing using Illumina MiSeq. Results The concentrations of 21 metabolites were significantly higher in the high exposure group than in the low exposure group(variable importance for projection>1,P<0.05).Most of these metabolites were positively correlated with the hydroxyl metabolites of naphthalene,fluorine,and phenanthrene(r=0.336-0.531).The identified differential metabolites primarily belonged to pathways associated with inflammation or proinflammatory states,including amino acid,lipid,and nucleotide metabolism.Additionally,these distinct metabolites were significantly associated with specific intestinal flora abundances(r=0.34-0.55),which were mainly involved in neurodevelopment. Conclusion Higher PAH exposure in young children affected metabolic homeostasis,particularly that of certain gut microbiota-derived metabolites.Further investigation is needed to explore the potential influence of PAHs on the gut microbiota and their possible association with neurodevelopmental outcomes.

Objective: This cross-sectional investigation aimed to determine the incidence, clinical characteristics, prognosis, and related risk factors of olfactory and gustatory dysfunctions related to infection with the SARS-CoV-2 Omicron strain in mainland China. Methods: Data of patients with SARS-CoV-2 from December 28, 2022, to February 21, 2023, were collected through online and offline questionnaires from 45 tertiary hospitals and one center for disease control and prevention in mainland China. The questionnaire included demographic information, previous health history, smoking and alcohol drinking, SARS-CoV-2 vaccination, olfactory and gustatory function before and after infection, other symptoms after infection, as well as the duration and improvement of olfactory and gustatory dysfunction. The self-reported olfactory and gustatory functions of patients were evaluated using the Olfactory VAS scale and Gustatory VAS scale. Results: A total of 35 566 valid questionnaires were obtained, revealing a high incidence of olfactory and taste dysfunctions related to infection with the SARS-CoV-2 Omicron strain (67.75%). Females(χ2=367.013, P<0.001) and young people(χ2=120.210, P<0.001) were more likely to develop these dysfunctions. Gender(OR=1.564, 95%CI: 1.487-1.645), SARS-CoV-2 vaccination status (OR=1.334, 95%CI: 1.164-1.530), oral health status (OR=0.881, 95%CI: 0.839-0.926), smoking history (OR=1.152, 95%CI=1.080-1.229), and drinking history (OR=0.854, 95%CI: 0.785-0.928) were correlated with the occurrence of olfactory and taste dysfunctions related to SARS-CoV-2(above P<0.001). 44.62% (4 391/9 840) of the patients who had not recovered their sense of smell and taste also suffered from nasal congestion, runny nose, and 32.62% (3 210/9 840) suffered from dry mouth and sore throat. The improvement of olfactory and taste functions was correlated with the persistence of accompanying symptoms(χ2=10.873, P=0.001). The average score of olfactory and taste VAS scale was 8.41 and 8.51 respectively before SARS-CoV-2 infection, but decreased to3.69 and 4.29 respectively after SARS-CoV-2 infection, and recovered to 5.83and 6.55 respectively at the time of the survey. The median duration of olfactory and gustatory dysfunctions was 15 days and 12 days, respectively, with 0.5% (121/24 096) of patients experiencing these dysfunctions for more than 28 days. The overall self-reported improvement rate of smell and taste dysfunctions was 59.16% (14 256/24 096). Gender(OR=0.893, 95%CI: 0.839-0.951), SARS-CoV-2 vaccination status (OR=1.334, 95%CI: 1.164-1.530), history of head and facial trauma(OR=1.180, 95%CI: 1.036-1.344, P=0.013), nose (OR=1.104, 95%CI: 1.042-1.171, P=0.001) and oral (OR=1.162, 95%CI: 1.096-1.233) health status, smoking history(OR=0.765, 95%CI: 0.709-0.825), and the persistence of accompanying symptoms (OR=0.359, 95%CI: 0.332-0.388) were correlated with the recovery of olfactory and taste dysfunctions related to SARS-CoV-2 (above P<0.001 except for the indicated values). Conclusion: The incidence of olfactory and taste dysfunctions related to infection with the SARS-CoV-2 Omicron strain is high in mainland China, with females and young people more likely to develop these dysfunctions. Active and effective intervention measures may be required for cases that persist for a long time. The recovery of olfactory and taste functions is influenced by several factors, including gender, SARS-CoV-2 vaccination status, history of head and facial trauma, nasal and oral health status, smoking history, and persistence of accompanying symptoms.
Female ; Humans ; Adolescent ; SARS-CoV-2 ; Smell ; COVID-19/complications* ; Cross-Sectional Studies ; COVID-19 Vaccines ; Incidence ; Olfaction Disorders/etiology* ; Taste Disorders/etiology* ; Prognosis

Objective:To explore the value of pre-ablation stimulated thyroglobulin (psTg) before 131I treatment combined with lymph node ratio (LNR) in predicting 131I treatment response in patients with papillary thyroid cancer (PTC). Methods:From January 2016 to December 2018, 178 PTC patients (47 males, 131 females; age (43.2±12.6) years) treated with 131I in the Affiliated Cancer Hospital of Zhengzhou University were retrospectively analyzed. According to 131I treatment response, patients were divided into excellent response (ER) group and non-ER group. The clinical data of the two groups were compared by χ2 test, independent-sample t test and Mann-Whitney U test. The cut-off values and AUCs of psTg and LNR to predict treatment response were calculated according to the ROC curve. Factors affecting 131I treatment response were analyzed by logistic multivariate regression analysis. Results:There were 118 patients (66.3%, 118/178) in ER group and 60 patients (33.7%, 60/178) in non-ER group, and there were significant differences in N stage ( χ2=11.15, P=0.004), 131I treatment dose ( χ2=12.65, P<0.001), American Thyroid Association (ATA) initial risk stratification ( χ2=15.25, P<0.001), number of metastatic lymph nodes ( χ2=22.63, P<0.001), LNR ( U=1 506.00, P<0.001) and psTg ( U=919.00, P<0.001) between the two groups. The cut-off values of psTg and LNR predicting ER were 3.97 μg/L and 0.29, with the AUC of 0.870 and 0.787 respectively. PsTg (odds ratio ( OR)=10.88, 95% CI: 4.67-25.36, P<0.001) and LNR ( OR=5.30, 95% CI: 1.85-15.23, P=0.002) were independent factors to predict 131I treatment response in PTC patients. When psTg≥3.97 μg/L, LNR ( OR=9.40, 95% CI: 2.06-42.92, P=0.004) was an independent factor affecting 131I treatment response in PTC patients. Conclusions:PsTg and LNR are independent factors affecting 131I treatment response in PTC patients. When psTg≥3.97 μg/L, LNR can be used as a supplementary factor to predict 131I treatment response. The combination of psTg and LNR can better predict 131I treatment response in PTC patients.

ObjectiveThe objective is to investigate the possibility of isocenter dual-guided resetting of surface guided radiation therapy （SGRT） combined with image guided radiation therapy （IGRT） in postoperative radiotherapy for breast cancer. To assess the setup error accuracy between the new resetting mode and the traditional resetting mode. MethodsRetrospective analysis was performed on breast cancer patients who underwent ELEKTA infinity accelerator radiotherapy in sun yat-sen university cancer center from July 13， 2021 to October 15， 2022. According to different reset methods， the patients were divided into a simulation group （41 cases） and a dual-guided group （40 cases）. The simulation group was reset using a simulator， CBCT scans were performed and setup errors were recorded during the first treatment； The dual-guided group was guided by AlignRT and combined with CBCT for isocenter dual-guided resetting， and the setup error obtained by CBCT registration was recorded. The global setup errors of chest region of interest （CROI） ， the local residual errors of supraclavicular region of interest （SROI） and the resetting time of the two reset methods were calculated and compared respectively. The advantages of the CBCT error distribution in the dual-guided resetting of SGRT combined with IGRT were analyzed. ResultsThe median of the global setup errors （X/cm， Y/cm， Z/cm， Rx°， Ry°， Rz°） of the simulation group and the median of the dual-guided group in the CROI were statistically significant （P<0.05） except the Rz and Ry directions. The local residual errors of the two groups of the SROI were calculated. The median of the errors of X/cm， Y/cm， Z/cm， Rx°， Ry°， Rz° were statistically significant （P<0.05） except the X and Y axis. The resetting time of the simulation group was significantly longer than that of the dual-guided group （238.64±28.56） s， t=-24.555， P=0.000， and the difference was statistically significant （P<0.05）. The CBCT error distribution of the dual-guide group was analyzed， and it was found that the absolute values of translation errors of X， Y and Z axis were all within 0.4 cm， while the proportions of ≤ 0.3 cm were 95%， 93% and 93%， respectively. The proportions of rotation errors of Rx， Ry and Rz ≤ 1.5 ° were 90%， 93% and 90%， respectively. ConclusionIn postoperative radiotherapy of breast cancer， SGRT combined with IGRT for isocenter dual-guided resetting can effectively correct the rotational setup errors and residual errors， and improve the accuracy of radiotherapy with less resetting time and high feasibility， which compared with the traditional simulator resetting mode. This precise， unmarked resetting method can be widely used in clinical practice.