Weiyong Liu

In the information era, the limited historical diagnostic data of radiologists are usually available. Under the conditions, how to help improve their diagnostic consistency and precision is an important issue. To address this issue, this study proposes an active learning-based data augmentation approach using unlabeled data, including data augmentation and best regression model hierarchies. In the data augmentation hierarchy, data-driven evidential reasoning (ER) models are constructed with an improved ER rule, in which criterion weights and reliabilities are learned from training data by giving two assumptions. These models rank unlabeled data in descending order according to informativeness and representativeness. A data-adding process and a label-giving process are then designed to add the appropriate pieces of unlabeled data to clusters of original training data and give labels to the added data to form augmented training data, respectively. In the best regression model hierarchy, the augmented training data and the testing data are used to determine the optimal model by using eight machine learning algorithms, which aims to improve consistent diagnostic capabilities of radiologists. In consideration of consistent diagnostic capabilities of radiologists, recommendations are designed to help improve their precise diagnostic capabilities. With the historical examination reports of the 16 radiologists in the two tertiary hospitals located in Hefei, Anhui, China, experiments are conducted to reveal that their consistent diagnostic capabilities are improved by the data augmentation approach. Appropriate recommendations are also generated to help improve their precise diagnostic capabilities. By discussing the managerial insights of the data augmentation approach and the designed recommendations, they are validated to be useful to help improve consistent and precise decision capabilities in management scenarios.

Glioblastoma (GBM) remains difficult to treat because the restrictive blood–brain barrier (BBB), the hypoxic tumor microenvironments, and mitochondrial electron transport chain (ETC)‐driven metabolic adaptability jointly limit therapeutic delivery and efficacy. Selective disruption of mitochondrial metabolism in GBM is challenging because ETC function is also essential for normal brain cells, and many therapeutic agents show limited BBB penetration or reduced activity under oxygen‐limited conditions. Here, we report Ni/LDH@M, an ultrasound‐responsive biomimetic Schottky nanoplatform assembled from nickel nanosheets and layered double hydroxide (LDH) nanosheets and cloaked with GL261 glioma cell membranes to enable homotypic recognition and accumulation in orthotopic glioma. Upon ultrasound irradiation, Ni/LDH@M drives carbon monoxide (CO) generation through interfacial electron modulation, and the generated CO inhibits mitochondrial cytochrome c oxidase, disrupts ETC electron transfer, and induces mitochondria‐dependent apoptosis and immunogenic cell death (ICD). In mice bearing orthotopic GL261 gliomas, ultrasound‐activated Ni/LDH@M suppressed tumor progression and prolonged survival, showing that controlled CO generation can reprogram tumor‐cell mitochondrial metabolism for GBM therapy.

To determine the diagnostic value of high-frequency ultrasonography in the assessment of anterior talofibular ligament (ATFL) injury and concomitant lesion

Objective. Ultrasound is the primary screening test for breast cancer. However, providing an interpretable auxiliary diagnosis of breast lesions is a challenging task. This study aims to develop an interpretable auxiliary diagnostic method to enhance usability in human-machine collaborative diagnosis. Approach. To address this issue, this study proposes the deep multi-stage reasoning method (DMSRM), which provides individual and overall breast imaging-reporting and data system (BI-RADS) assessment categories for breast lesions. In the first stage of the DMSRM, the individual BI-RADS assessment network (IBRANet) is designed to capture lesion features from breast ultrasound images. IBRANet performs individual BI-RADS assessments of breast lesions using ultrasound images, focusing on specific features such as margin, contour, echogenicity, calcification, and vascularity. In the second stage, evidence reasoning (ER) is employed to achieve uncertain information fusion and reach an overall BI-RADS assessment of the breast lesions. Main results. To evaluate the performance of DMSRM at each stage, two test sets are utilized: the first for individual BI-RADS assessment, containing 4322 ultrasound images; the second for overall BI-RADS assessment, containing 175 sets of ultrasound image pairs. In the individual BI-RADS assessment of margin, contour, echogenicity, calcification, and vascularity, IBRANet achieves accuracies of 0.9491, 0.9466, 0.9293, 0.9234, and 0.9625, respectively. In the overall BI-RADS assessment of lesions, the ER achieves an accuracy of 0.8502. Compared to independent diagnosis, the human-machine collaborative diagnosis results of three radiologists show increases in positive predictive value by 0.0158, 0.0427, and 0.0401, in sensitivity by 0.0400, 0.0600 and 0.0434, and in area under the curve by 0.0344, 0.0468, and 0.0255. Significance. This study proposes a DMSRM that enhances the transparency of the diagnostic reasoning process. Results indicate that DMSRM exhibits robust BI-RADS assessment capabilities and provides an interpretable reasoning process that better suits clinical needs.

Objectives: Our study aims to investigate the impact of B‐mode ultrasound (B‐US) imaging, color Doppler flow imaging (CDFI), strain elastography (SE), and patient age on the prediction of molecular subtypes in breast lesions.Methods: Totally 2272 multimodal ultrasound imaging was collected from 198 patients. The ResNet‐18 network was employed to predict four molecular subtypes from B‐US imaging, CDFI, and SE of patients with different ages. All the images were split into training and testing datasets by the ratio of 80%:20%. The predictive performance on testing dataset was evaluated through 5 metrics including mean accuracy, precision, recall, F1‐scores, and confusion matrix.Results: Based on B‐US imaging, the test mean accuracy is 74.50%, the precision is 74.84%, the recall is 72.48%, and the F1‐scores is 0.73. By combining B‐US imaging with CDFI, the results were increased to 85.41%, 85.03%, 85.05%, and 0.84, respectively. With the integration of B‐US imaging and SE, the results were changed to 75.64%, 74.69%, 73.86%, and 0.74, respectively. Using images from patients under 40 years old, the results were 90.48%, 90.88%, 88.47%, and 0.89. When images from patients who are above 40 years old, they were changed to 81.96%, 83.12%, 80.5%, and 0.81, respectively.Conclusion: Multimodal ultrasound imaging can be used to accurately predict the molecular subtypes of breast lesions. In addition to B‐US imaging, CDFI rather than SE contribute further to improve predictive performance. The predictive performance is notably better for patients under 40 years old compared with those who are 40 years old and above.

Background and Aims. Intestinal ultrasound (IUS) is considered a nonirradiating, noninvasive, well-tolerated, and valuable tool for objectively assessing Crohn’s disease (CD) activity. However, there is no widely accepted intestinal ultrasound scoring system. This study is aimed at evaluating the efficacy of IUS key parameters, the International Bowel Ultrasound Activity Score (IBUS-SAS), and IBUS-SAS combined with blood inflammatory markers in assessing CD activity. Methods. 40 CD patients were reviewed in this retrospective study and were divided into the moderate-severe group ( ) and nonmoderate-severe group ( ) based on a simplified endoscopic score of Crohn’s disease (SES-CD). Double-balloon enteroscopy/colonoscopy were reviewed by three gastroenterologists. A transabdominal ultrasound was performed by two ultrasound specialists. Blood inflammatory markers were measured from morning samples. Results. In evaluating moderate to severe CD patients, (1) IBUS-SAS had a good predictive effect with an area-under-the-curve (AUC) of 0.855 ( ); (2) IUS key parameters (including BWT, CDS, BWS, and I-fat) yielded good predictive effects with AUC of 0.811, 0.731, 0.724, and 0.747, respectively ( ); (3) blood inflammatory markers (including ESR, CRP, PLR, MLR, and NLR) also had good predictive effects with AUC of 0.771, 0.837, 0.728, 0.743, and 0.775, respectively ( ); (4) IBUS-SAS combined with ESR and CRP exerted the best predictive effect with the highest AUC of 0.912 (95% CI: 0.823-1.000), and the sensitivity and specificity were 88.0% and 80.0%, respectively ( ). Conclusion. IBUS-SAS combined with ESR and CRP is a more efficient tool than IBUS-SAS alone or inflammatory markers alone in evaluating CD patients with moderate to severe disease activity.

Background: Easily accessible tools that reliably stratify Mycobacterium tuberculosis (MTB) infection are needed to facilitate the improvement of clinical management. The current study attempts to reveal lymphocyte-related immune characteristics of active tuberculosis (ATB) patients and establish immunodiagnostic model for discriminating ATB from latent tuberculosis infection (LTBI) and healthy controls (HC).Methods: A total of 171 subjects consisted of 54 ATB, 57 LTBI, and 60 HC were consecutively recruited at Tongji hospital from January 2019 to January 2021. All participants were tested for lymphocyte subsets, phenotype, and function. Other examination including T-SPOT and microbiological detection for MTB were performed simultaneously.Results: Compared with LTBI and HC, ATB patients exhibited significantly lower number and function of lymphocytes including CD4⁺ T cells, CD8⁺ T cells and NK cells, and significantly higher T cell activation represented by HLA-DR and proportion of immunosuppressive cells represented by Treg. An immunodiagnostic model based on the combination of NK cell number, HLA-DR⁺CD3⁺ T cells, Treg, CD4⁺ T cell function, and NK cell function was built using logistic regression. Based on receiver operating characteristic curve analysis, the area under the curve (AUC) of the diagnostic model was 0.920 (95% CI, 0.867-0.973) in distinguishing ATB from LTBI, while the cut-off value of 0.676 produced a sensitivity of 81.48% (95% CI, 69.16%-89.62%) and specificity of 91.23% (95% CI, 81.06%-96.20%). Meanwhile, AUC analysis between ATB and HC according to the diagnostic model was 0.911 (95% CI, 0.855-0.967), with a sensitivity of 81.48% (95% CI, 69.16%-89.62%) and a specificity of 90.00% (95% CI, 79.85%-95.34%).Conclusions: Our study demonstrated that the immunodiagnostic model established by the combination of lymphocyte-related indicators could facilitate the status differentiation of MTB infection.