Search

To shorten the wet-extraction process of vanadium slag and reduce the loss of a large number of valuable elements—such as iron, manganese, and chromium—this study directly used molten vanadium slag to smelt alloys. Vanadium slag with 16 wt.% carbon, 20 wt.% CaO (R = 0.97), and 5 wt.% Al2O3 was reduced at 1550°C for 2 h, and a Fe-V-(Cr-Mn) alloy with Fe-V as the matrix was obtained, which could be used as a raw material for vanadium alloy steel. The valuable elements of iron, vanadium, and chromium in the slag were almost completely reduced. The tailing was mainly composed of Ca3Si3O9, Ca(Mg,Mn)(Si,Al)2O6, and CaAl2SiO6, which are primarily silicates and aluminates. To obtain a better reduction rate of elements, the appropriate reduction temperature, reduction agent, and modification agent were 1500°C, 16–20 wt.% C, and 20 wt.% CaO (R = 0.97), respectively. In addition, by effectively controlling the process parameters, the recovery rate of iron and vanadium could reach as high as 98%, and the recovery rate of chromium and manganese could reach > 90% and 70%, respectively, which could effectively recover valuable elements in vanadium slag and the sensible heat, shorten the vanadium extraction process, and decrease the amount of the three waste types. [ABSTRACT FROM AUTHOR]

Purpose: Mice can develop arterial damage and even atherosclerosis under intermittent hypoxia (IH); however, the specific mechanism of arterial damage induced by IH remains unclear. Hence, this research aimed to illustrate the underlying mechanism linking IH to arterial injury. Materials and methods: The differential gene expression of the thoracic aorta under normoxia or IH mice was analyzed utilizing RNA sequencing. Furthermore, GO, KEGG pathway, and CIBERSORT analyses were carried out. For verification of the expression of candidate genes affected by IH, quantitative RT-qPCR (qRT-PCR) was conducted. Immunohistochemical (IHC) staining revealed immune cell infiltration in the thoracic aorta. Results: The thickness of the intima-media of the mouse aorta was increased, and the fiber structure was disordered under IH. Transcriptomics analysis showed that in the aorta, 1137 upregulated genes and 707 downregulated genes were affected by IH, significantly related to the activation of the immune system and cell adhesion. Furthermore, B cell infiltration around the aorta was observed under IH. Conclusions: IH might lead to structural changes in the aorta by activating the immune response and enhancing cell adhesion. [ABSTRACT FROM AUTHOR]

Atherosclerotic renovascular disease (RVD) leads to hypertension, chronic kidney disease (CKD), and heart disease. Intrarenal delivery of mesenchymal stem cells (MSCs) and MSC-derived extracellular vesicles (EVs) attenuate renal injury and suppress release of inflammatory cytokines in porcine RVD. We hypothesized that this strategy would also be useful for cardioprotection. Pigs with renovascular hypertension and metabolic syndrome were studied 4 weeks after treatment with a single intrarenal infusion of autologous MSCs, EVs, or vehicle. Cardiac structure and function were assessed in vivo, and myocardial remodeling and expression of the pro-fibrotic factor growth factor receptor-bound protein-2 (Grb2) were measured ex-vivo. Inflammatory cytokine levels were measured in the systemic circulation and myocardial tissue. Blood pressure was elevated in all RVD groups, but serum creatinine increased in RVD and decreased in both RVD + MSCs and RVD + EVs. RVD-induced diastolic dysfunction (lower E/A ratio) was normalized in both MSCs- and EVs- treated pigs. Intrarenal delivery of MSCs and EVs also attenuated RVD-induced myocardial fibrosis, collagen deposition, and Grb2 expression, yet EVs restored capillary density and inflammation more effectively than MSCs. These observations suggest that autologous EVs attenuate cardiac injury in experimental RVD more effectively than their parent MSCs. [ABSTRACT FROM AUTHOR]

Objective: To determine if artificial intelligence–based modification of the Thyroid Imaging Reporting Data System (TI-RADS) would be better than the current American College of Radiology (ACR) TI-RADS for risk stratification of thyroid nodules. Methods: A total of 2061 thyroid nodules (in 1859 patients) sampled with fine-needle aspiration or operation were retrospectively analyzed between January 2017 and July 2020. Two radiologists blinded to the pathologic diagnosis evaluated nodule features in five ultrasound categories and assigned TI-RADS scores by both ACR TI-RADS and AI TI-RADS. Inter-rater agreement was assessed by asking another two radiologists to score a set of 100 nodules independently. The reference standard was postoperative pathological or cytopathological diagnosis according to the Bethesda system. Inter-rater agreement was determined using intraclass correlation coefficient (ICC). Results: AI TI-RADS assigned lower TI-RADS risk levels than ACR TI-RADS (p < 0.001) and had larger area under receiver operating characteristic curve (0.762 vs. 0.679, p < 0.001). The sensitivities of ACR TI-RADS and AI TI-RADS were similar (86.7% vs. 82.2%, p = 0.052), but specificity was higher with AI TI-RADS (70.2% vs. 49.2%, p < 0.001). AI TI-RADS downgraded 743 (48.63%) benign nodules, indicating that 328 (42.3% of 776 biopsied nodules) unnecessary fine-needle aspirations (FNA) could have been avoided. Inter-rater agreement was better with AI TI-RADS than with ACR TI-RADS (ICC, 0.808 vs. 0.861, p < 0.001). Conclusion: AI TI-RADS can achieve meaningful reduction in the number of benign thyroid nodules recommended for biopsy and significantly improve specificity despite a slight decrease in sensitivity. Key Points: • AI TI-RADS assigned lower TI-RADS risk levels than ACR TI-RADS, showing similar sensitivity but higher specificity. • Half of the benign nodules can be downgraded of which 42.3% of biopsy nodules avoided unnecessary fine-needle aspiration (FNA). • AI TI-RADS had a better overall inter-rater agreement. [ABSTRACT FROM AUTHOR]

Cellular senescence is a ubiquitous process with roles in tissue remodelling, including wound repair and embryogenesis. However, prolonged senescence can be maladaptive, leading to cancer development and age-related diseases. Cellular senescence involves cell-cycle arrest and the release of inflammatory cytokines with autocrine, paracrine and endocrine activities. Senescent cells also exhibit morphological alterations, including flattened cell bodies, vacuolization and granularity in the cytoplasm and abnormal organelles. Several biomarkers of cellular senescence have been identified, including SA-βgal, p16 and p21; however, few markers have high sensitivity and specificity. In addition to driving ageing, senescence of immune and parenchymal cells contributes to the development of a variety of diseases and metabolic disorders. In the kidney, senescence might have beneficial roles during development and recovery from injury, but can also contribute to the progression of acute kidney injury and chronic kidney disease. Therapies that target senescence, including senolytic and senomorphic drugs, stem cell therapies and other interventions, have been shown to extend lifespan and reduce tissue injury in various animal models. Early clinical trials confirm that senotherapeutic approaches could be beneficial in human disease. However, larger clinical trials are needed to translate these approaches to patient care. [ABSTRACT FROM AUTHOR]

Purpose: Glutamine plays an important role in tumor metabolism and progression. This research aimed to find out how Gln exert their effects on laryngeal squamous cell carcinoma (LSCC). Methods: Cell proliferation was measured by CCK8 and EdU assay, mitochondrial bioenergetic activity was measured by mitochondrial stress tests. Gene expression profiling was revealed by RNA sequencing and validated by RT-qPCR. In LSCC patients, protein expression in tumor and adjacent tissues was examined and scored by IHC staining. RNAi was performed by stably expressed shRNA in TU177 cells. In vivo tumor growth analysis was performed using a nude mouse tumorigenicity model. Results: Gln deprivation suppressed TU177 cell proliferation, which was restored by αKG supplementation. By transcriptomic analysis, we identified CECR2, which encodes a histone acetyl-lysine reader, as the downstream target gene for Gln and αKG. In LSCC patients, the expression of CECR2 in tumors was lower than adjacent tissues. Furthermore, deficiency of CECR2 promoted tumor cell growth both in vitro and in vivo, suggesting it has tumor suppressor effects. Besides, cell proliferation inhibited by Gln withdrawal could be restored by CECR2 depletion, and the proliferation boosted by αKG supplementation could be magnified either, suggested that CECR2 feedback suppressed Gln and αKG's effect on tumor growth. Transcriptomic profiling revealed CECR2 regulated the expression of a series of genes involved in tumor progression. Conclusion: We confirmed the Gln-αKG-CECR2 axis contributes to tumor growth in LSCC. This finding provided a potential therapeutic opportunity for the use of associated metabolites as a potential treatment for LSCC. [ABSTRACT FROM AUTHOR]

Premature ovarian insufficiency (POI) is defined as the loss of ovarian activity under the age of 40. Theca cells (TCs) play a vital role during folliculogenesis and TCs dysfunction participate in the pathogenesis of POI. Therefore, transplantation of thecal stem cells (TSCs), which are capable of self-renewal and differentiation into mature TCs, may provide a new strategy for treating POI. To investigate the feasibility, safety, and efficacy of TSCs transplantation in clinically relevant non-human primate (NHP) models, we isolate TSCs from cynomolgus monkeys, and these cells are confirmed to expand continuously and show potential to differentiate into mature TCs. In addition, engraftment of autologous TSCs into POI monkeys significantly improves hormone levels, rescues the follicle development, promotes the quality of oocytes and boosts oocyte maturation/fertilization rate. Taken together, these results for the first time suggest that autologous TSCs can ameliorate POI symptoms in primate models and shed new light on developing stem cell therapy for POI. [ABSTRACT FROM AUTHOR]

Purpose: To evaluate whether a predictive model based on nocturnal minimal oxygen saturation (SpO2) alone can accurately detect the presence of obstructive sleep apnea (OSA) in a population with suspected OSA. Methods: A total of 4297 participants with suspected OSA were enrolled in this study, and laboratory-based polysomnography (PSG) tests were performed at sea level in all subjects. Nocturnal minimal SpO2 was obtained automatically as part of the PSG test. Stratified sampling was used to divide the participants' data into the training set (75%) and the test set (25%). An OSA detection model based on minimal SpO2 alone was created using the training set data and its performance was evaluated using the independent test set data ("hold-out" evaluation). Gender-specific models, and models based on minimal SpO2 in combination with other predictive factors (age, body mass index, waist-to-hip ratio, snoring grade, Epworth Sleepiness Scale score, and comorbidities), were also created and compared in terms of OSA detection performance. Results: The prevalence of OSA was 85.6% in our study population. The models including multiple predictors, and the gender-specific models, failed to outperform the model based solely on minimal SpO2, which showed good predictive performance (C statistic, 0.922) having an overall accuracy rate of 0.86, sensitivity of 0.87, specificity of 0.84, positive predictive value of 0.97, and positive likelihood ratio of 5.34. In addition, the model based on minimal SpO2 alone could also accurately predict the presence of moderate-to-severe OSA and severe OSA, with C statistics of 0.914 and 0.900, respectively. Conclusions: A predictive model based on nocturnal minimal SpO2 alone may be an alternative option to detect the presence of OSA in a high-risk population when standard diagnostic tests are unavailable. [ABSTRACT FROM AUTHOR]

Eye-control interfaces are human–computer interfaces in which interaction is mediated by the user's gaze. Using dwell time for target selection may be hindered by the "Midas Touch" problem, which posits that intentional selection and perceptual processes cannot be separated. To solve this problem, we investigated the influence of different dwell times on task performance. Results suggest that the optimal dwell time to trigger a click in eye-control movement plus eye-control click and hand-control movement plus eye-control click are 700 and 200 ms, respectively. In addition, the eye-control movement plus eye-control click mode has a lower completion rate than the hand-control movement plus eye-control click mode. [ABSTRACT FROM AUTHOR]

Obesity is strongly correlated with obstructive sleep apnea (OSA), and bariatric surgery can effectively treat obesity and alleviate OSA. However, the contributing factors are still unclear. We aimed to explore the relationship between betatrophin and OSA in patients undergoing Roux-en-Y gastric bypass (RYGB) surgery. Our study consisted of thirty-seven individuals with OSA and type 2 diabetes (16 males, 21 females) undergoing RYGB surgery. The polysomnography test, anthropometric results, serum betatrophin, and abdominal magnetic resonance images were evaluated both before and 1 year after RYGB surgery. Factors that may correlate with the alleviation of OSA were investigated. In our study, RYGB surgery significantly decreased apnea hypopnea index (AHI) and serum betatrophin concentration (p < 0.001). The abdominal visceral fat area, subcutaneous fat area and HOMA-IR were also significantly decreased (p < 0.001). The preoperative AHI, postoperative AHI and the change in AHI were significantly correlated with the preoperative betatrophin, postoperative betatrophin and the change in betatrophin, respectively (p < 0.05). These correlations were still significant after adjustment for other risk factors. The change in betatrophin was also independently associated with the change in minimum oxygen saturation (p < 0.001). Our data might indicate that serum betatrophin was significantly independently correlated with the improvement of OSA after bariatric surgery. [ABSTRACT FROM AUTHOR]

This paper reveals two symmetry properties of the states and costates of a special type of optimal impulsive control problems and, based on these two properties, proposes a new procedure for obtaining the second-order states and costates approximations. It further shows with an example that this procedure is efficient and can be implemented effectively in a computer algebraic system. [ABSTRACT FROM AUTHOR]

Computer-aided process planning is an important interface between computer-aided design and computer-aided manufacturing in computer-integrated manufacturing environments. In this paper, the complicated process planning is modeled as a combinatorial optimization problem with constraints, and a hybrid graph and genetic algorithm (GA) approach has been developed. The approach deals with process planning problems in a concurrent manner by simultaneously considering activities such as sequencing operations, selecting manufacturing resources, and determining setup plans to achieve the global optimal objective. Graph theory accompanied with matrix theory, as the basic mathematical tool for operation sequencing, is embedded into the main frame of GA. The precedence constraints between operations are formulated in an operation precedence graph (OPG). The initial population composed of all feasible solutions is generated by an elaborately designed topologic sort algorithm to the OPG. A modified crossover operator guaranteeing only feasible offspring generated is used, two types of mutation strategies are adopted, and a heuristic algorithm is applied to adjust the infeasible plan generated by the mutation operator to the feasible domain. A case study has been carried out to demonstrate the feasibility and efficiency of the proposed approach. [ABSTRACT FROM AUTHOR]

Trajectory planning for satellite formation flying missions requires the ability to find the optimal control law to transfer a satellite from one periodic relative orbit to another. This article modifies Jezewski's linear impulsive trajectory optimization method for solving nonlinear problems described in the relative coordinate frame, and proposes the concept of standard problems for the initialization of the optimizer. With this modified optimizer and the standard problem concept, the article studies the impulse number, the cost, and the nonlinearity of fixed-time coplanar transfer problems. [ABSTRACT FROM AUTHOR]

Infantile malignant osteopetrosis (IMO) (OMIM 259700) is a lethal autosomal recessive disease. The underlying gene in most IMO patients is TCIRG1. This codes for the TCIRG1 protein involved in the cellular proton pump, which is highly expressed on surfaces of osteoclasts. We have characterized a family comprising two affected siblings born to healthy parents. The sister and her younger brother both presented classical X-ray images of IMO at 17 h and 16 weeks, respectively, after birth, and both died after the appearance of fever and flu-like symptoms months later. Radiographs revealed normal bone density in both parents. Mutation detection of the TCIRG1 gene was performed in the boy and the parents. The novel mutation c.242delC (p.Pro81ArgfsX85) and the known mutation c.1114C>T (p.Gln372X) were both identified in the boy. Both mutations are predicted to introduce premature stop codons, with deletion of 666 amino acids from the C terminus of the TCIRG1 protein of one allele and 459 from the other. Both mutations involve loss of part or the whole of the ATPase V0-complex domain of the protein. The father carries the c.242delC (p.Pro81ArgfsX85) mutation and the mother the c.1114C>T (p.Gln372X). Our findings provide new data for pre- and post-natal genetic diagnosis and identification of heterozygous carriers of the disease. [ABSTRACT FROM AUTHOR]

The 2009 flu pandemic is caused by a new strain of influenza A (H1N1) virus, A/H1N1/09. With its high transmissibility, this novel virus has caused a pandemic and infected over 600,000 people globally. By comparing the hemaglutinin (HA) gene and protein sequences among over 700 A/H1N1/09 isolates, mutations in the receptor-binding sites and antigenic epitope regions were identified. Among these mutations, T220 and E/G239 were found to be strongly positively selected over the course of spreading of the A/H1N1/09 virus worldwide. Interestingly, both sites are located in the highly variable epitope regions of HA1, and residue 239 also plays an important role in the receptor-binding process. Further analyses demonstrated that the percentage of T220 mutants among all isolates increased rapidly during the evolution, and that an E/G239 mutation could decrease the binding affinity of the virus with its cellular receptor. Thus, due to a potential functional importance of residues 220 and 239, mutations at these sites, as well as the significant of positive selection on these sites deserves more attention, while new vaccines and therapeutic drugs are developed against this novel virus. [ABSTRACT FROM AUTHOR]

Genomic variation is the genetic basis of phenotypic diversity among individuals, including variation in disease susceptibility and drug response. The greatest promise of the International HapMap is to provide roadmaps for identifying genetic variants predisposing to complex diseases. Single nucleotide polymorphism (SNP) is the fundamental element of the HapMap. Allele frequency of SNPs is one of the major factors affecting the resulting HapMap, being the factor upon which linkage disequilibrium (LD) is calculated, haplotypes are constructed, and tagging SNPs (tagSNPs) are selected. The cutoff thresholds for the frequency of minor alleles used in the making of the map therefore have profound effects on the resolution of that map. To date most researchers have adopted their own cutoff thresholds, and there has been little real dataset-based evaluation of the effects of different cutoff thresholds on HapMap resolution. In an attempt to assess the implications of different cutoff values, we analyzed our own data for the centromeric genes on Chromosome 15 in Chinese Han and Tibetan populations, with respect to minor allele frequency cutoff values of ⩾0.01 (0.01 group), ⩾0.05 (0.05 group), and ⩾0.10 (0.10 group), and constructed HapMaps from each of the datasets. The resolution, study power and cost-effectiveness for each of the maps were compared. Our results show that the 0.01 threshold provides the greatest power (P= 0.019 in Han and P= 0.029 in Tibetan for 0.01 vs. 0.05 threshold) and detects most population-specific haploypes (P= 0.012 for 0.01 vs. 0.05 threshold). However, in the regions studied, the 0.05 cutoff threshold did not significantly increase power above the 0.10 threshold (P = 0.191 in Han; 1.000 in Tibetans), and did not improve resolution over the 0.10 value for population- specific haplotypes (P= 0.592) neither. Furthermore the 0.05 and 0.10 values produced the same figures for tagging efficiency, LD block number, LD length, study power and cost-savings in the Tibetan population. These results suggest that a lower cutoff value is more appropriate for studies in which population-specific haplotypes are crucial, and that the most appropriate cutoff value may differ between populations. Due to the limited genes studied in this project more studies should be conducted to further address this important issue. [ABSTRACT FROM AUTHOR]

Genetic variations and their functional implications have been one of the focuses in recent genome research. With the release of the HapMap by the International Consortium, and the availability of the ultra-high-volume genotyping platform, it will soon be possible to use genome-wide association approach to identify genetic variations responsible for complex traits/diseases. While the power of this approach is generally agreed, it is a debated issue as to how much population difference should be exploited, and how best it should be applied. To address this issue we have sequenced 7 genes in the centromeric region of chromosome 15, investigated their SNPs, SNP frequencies, tagSNPs, LD structures, and haplotypes in 50 Tibetan subjects, and compared them with those from the Han population. Genetic diversities between the two populations were also quantified. Our results show that the overall genetic variation between the two populations is very little, but there are differences, primarily in allele frequencies, which is a dominating factor for haplotypes and tagSNPs. In general Tibetans have longer LD and less diversity in the region studied. These data provide genetic evidence for the close relationship between the two populations, and support the idea that all populations are fundamentally the same, but also indicate population variations, particularly in allele frequency, should be taken into account in complex traits/diseases analysis. Data obtained in this investigation not only help us understand the genome region, but also provide road maps for variation study in the genes/ region in Tibetan population. [ABSTRACT FROM AUTHOR]

Recent advances have shown that the majority of the nucleotide variation in human genome is single nucleotide polymorphisms (SNPs). Using SNPs each chromosome can be divided into different haplotype blocks, and there are limited common haplotypes in each block. This provides a powerful approach for whole genome scan for disease-associated genes/variants. However, most data available today are based on the large-scale genomic analyses, data concerning individual genes for fine mapping with high density SNPs are relatively lacking. We have sequenced 7 genes and their flanking regions, identified 34 novel SNPs, constructed high density SNP haplotypes and haplotype blocks in 5 genes in the centromeric region of chromosome 15 in 100 Chinese Han subjects. Our results show that there is a great heterogeneity in the haplotypes and haplotype block structures within and between these genes, which are in close physical proximity. Data obtained in this study provide a useful tool for candidate gene approach at the fine scale for identifying disease contributing variants in the genes/regions. [ABSTRACT FROM AUTHOR]

Pericytes play essential roles in blood–brain barrier (BBB) integrity and dysfunction or degeneration of pericytes is implicated in a set of neurological disorders although the underlying mechanism remains largely unknown. However, the scarcity of material sources hinders the application of BBB models in vitro for pathophysiological studies. Additionally, whether pericytes can be used to treat neurological disorders remains to be elucidated. Here, we generate pericyte-like cells (PCs) from human pluripotent stem cells (hPSCs) through the intermediate stage of the cranial neural crest (CNC) and reveal that the cranial neural crest-derived pericyte-like cells (hPSC-CNC PCs) express typical pericyte markers including PDGFRβ, CD146, NG2, CD13, Caldesmon, and Vimentin, and display distinct contractile properties, vasculogenic potential and endothelial barrier function. More importantly, when transplanted into a murine model of transient middle cerebral artery occlusion (tMCAO) with BBB disruption, hPSC-CNC PCs efficiently promote neurological functional recovery in tMCAO mice by reconstructing the BBB integrity and preventing of neuronal apoptosis. Our results indicate that hPSC-CNC PCs may represent an ideal cell source for the treatment of BBB dysfunction-related disorders and help to model the human BBB in vitro for the study of the pathogenesis of such neurological diseases. Pericytes play an essential role in blood brain barrier (BBB) integrity. Here, the authors generate pericyte-like cells (PCs) from human pluripotent stem cells (hPSCs) which display functional properties and also promote BBB recovery in a mouse model of cerebral artery occlusion. [ABSTRACT FROM AUTHOR]

Hereditary spastic paraplegia (HSP) is a group of neurodegenerative diseases. The genotypes and phenotypes of HSP are extremely heterogenous. SPG3A is one of the identified genes underlying HSP, and codes for a GTPase, atlastin. Mutations in SPG3A are currently believed to be associated with early onset and mild phenotypes. And most structural predictions could not detect gross changes in the mutant protein. However, in a severely affected HSP family we have identified a novel SPG3A mutation, c.1228G>A (p.G410R), in a Tibetan kindred. The mutation occurred at the highly conserved nucleotide and co-segregated with the disease, and was absent in the control subjects. Structural predictions showed that the Tibetan mutation occurred at the linking part between the guanylate-binding protein domain (GB, the ball region) and the transmembrane helices (TM, the rod region) at the start point of an a-helix, which may disrupt the helix, and cause changes in the overall structure of the transmembrane region of the molecule. Our results indicate that severe phenotypes can also arise from SPG3A mutations and the linking part of the guanylate-binding protein domain and the transmembrane helices might be crucial in determining the severity of the disease. This paper not only presents the first SPG3A mutational report from the Chinese population, but also provides potential evidence for a possible correlation between the severity of the phenotypes of HSP with the extension of the changes in the protein structures of atlastin. [ABSTRACT FROM AUTHOR]

Objectives: The preoperative classification of pleomorphic adenomas (PMA) and Warthin tumors (WT) in the parotid gland plays an essential role in determining therapeutic strategies. This study aims to develop and validate an ultrasound-based ensemble machine learning (USEML) model, employing nonradiative and noninvasive features to differentiate PMA from WT.A total of 203 patients with histologically confirmed PMA or WT who underwent parotidectomy from two centers were enrolled. Clinical factors, ultrasound (US) features, and radiomic features were extracted to develop three types of machine learning model: clinical models, US models, and USEML models. The diagnostic performance of the USEML model, as well as that of physicians based on experience, was evaluated and validated using receiver operating characteristic (ROC) curves in internal and external validation cohorts. DeLong’s test was used for comparisons of AUCs. SHAP values were also utilized to explain the classification model.The USEML model achieved the highest AUC of 0.891 (95% CI, 0.774–0.961), surpassing the AUCs of both the US (0.847; 95% CI, 0.720–0.932) and clinical (0.814; 95% CI, 0.682–0.908) models. The USEML model also outperformed physicians in both internal and external validation datasets (both p < 0.05). The sensitivity, specificity, negative predictive value, and positive predictive value of the USEML model and physician experience were 89.3%/75.0%, 87.5%/54.2%, 87.5%/65.6%, and 89.3%/65.0%, respectively.The USEML model, incorporating clinical factors, ultrasound factors, and radiomic features, demonstrated efficient performance in distinguishing PMA from WT in the parotid gland.This study developed a machine learning model for preoperative diagnosis of pleomorphic adenoma and Warthin tumor in the parotid gland based on clinical, ultrasound, and radiomic features. Furthermore, it outperformed physicians in an external validation dataset, indicating its potential for clinical application.• Differentiating pleomorphic adenoma (PMA) and Warthin tumor (WT) affects management decisions and is currently done by invasive biopsy.• Integration of US-radiomic, clinical, and ultrasound findings in a machine learning model results in improved diagnostic accuracy.• The ultrasound-based ensemble machine learning (USEML) model consistently outperforms physicians, suggesting its potential applicability in clinical settings.Methods: The preoperative classification of pleomorphic adenomas (PMA) and Warthin tumors (WT) in the parotid gland plays an essential role in determining therapeutic strategies. This study aims to develop and validate an ultrasound-based ensemble machine learning (USEML) model, employing nonradiative and noninvasive features to differentiate PMA from WT.A total of 203 patients with histologically confirmed PMA or WT who underwent parotidectomy from two centers were enrolled. Clinical factors, ultrasound (US) features, and radiomic features were extracted to develop three types of machine learning model: clinical models, US models, and USEML models. The diagnostic performance of the USEML model, as well as that of physicians based on experience, was evaluated and validated using receiver operating characteristic (ROC) curves in internal and external validation cohorts. DeLong’s test was used for comparisons of AUCs. SHAP values were also utilized to explain the classification model.The USEML model achieved the highest AUC of 0.891 (95% CI, 0.774–0.961), surpassing the AUCs of both the US (0.847; 95% CI, 0.720–0.932) and clinical (0.814; 95% CI, 0.682–0.908) models. The USEML model also outperformed physicians in both internal and external validation datasets (both p < 0.05). The sensitivity, specificity, negative predictive value, and positive predictive value of the USEML model and physician experience were 89.3%/75.0%, 87.5%/54.2%, 87.5%/65.6%, and 89.3%/65.0%, respectively.The USEML model, incorporating clinical factors, ultrasound factors, and radiomic features, demonstrated efficient performance in distinguishing PMA from WT in the parotid gland.This study developed a machine learning model for preoperative diagnosis of pleomorphic adenoma and Warthin tumor in the parotid gland based on clinical, ultrasound, and radiomic features. Furthermore, it outperformed physicians in an external validation dataset, indicating its potential for clinical application.• Differentiating pleomorphic adenoma (PMA) and Warthin tumor (WT) affects management decisions and is currently done by invasive biopsy.• Integration of US-radiomic, clinical, and ultrasound findings in a machine learning model results in improved diagnostic accuracy.• The ultrasound-based ensemble machine learning (USEML) model consistently outperforms physicians, suggesting its potential applicability in clinical settings.Results: The preoperative classification of pleomorphic adenomas (PMA) and Warthin tumors (WT) in the parotid gland plays an essential role in determining therapeutic strategies. This study aims to develop and validate an ultrasound-based ensemble machine learning (USEML) model, employing nonradiative and noninvasive features to differentiate PMA from WT.A total of 203 patients with histologically confirmed PMA or WT who underwent parotidectomy from two centers were enrolled. Clinical factors, ultrasound (US) features, and radiomic features were extracted to develop three types of machine learning model: clinical models, US models, and USEML models. The diagnostic performance of the USEML model, as well as that of physicians based on experience, was evaluated and validated using receiver operating characteristic (ROC) curves in internal and external validation cohorts. DeLong’s test was used for comparisons of AUCs. SHAP values were also utilized to explain the classification model.The USEML model achieved the highest AUC of 0.891 (95% CI, 0.774–0.961), surpassing the AUCs of both the US (0.847; 95% CI, 0.720–0.932) and clinical (0.814; 95% CI, 0.682–0.908) models. The USEML model also outperformed physicians in both internal and external validation datasets (both p < 0.05). The sensitivity, specificity, negative predictive value, and positive predictive value of the USEML model and physician experience were 89.3%/75.0%, 87.5%/54.2%, 87.5%/65.6%, and 89.3%/65.0%, respectively.The USEML model, incorporating clinical factors, ultrasound factors, and radiomic features, demonstrated efficient performance in distinguishing PMA from WT in the parotid gland.This study developed a machine learning model for preoperative diagnosis of pleomorphic adenoma and Warthin tumor in the parotid gland based on clinical, ultrasound, and radiomic features. Furthermore, it outperformed physicians in an external validation dataset, indicating its potential for clinical application.• Differentiating pleomorphic adenoma (PMA) and Warthin tumor (WT) affects management decisions and is currently done by invasive biopsy.• Integration of US-radiomic, clinical, and ultrasound findings in a machine learning model results in improved diagnostic accuracy.• The ultrasound-based ensemble machine learning (USEML) model consistently outperforms physicians, suggesting its potential applicability in clinical settings.Conclusions: The preoperative classification of pleomorphic adenomas (PMA) and Warthin tumors (WT) in the parotid gland plays an essential role in determining therapeutic strategies. This study aims to develop and validate an ultrasound-based ensemble machine learning (USEML) model, employing nonradiative and noninvasive features to differentiate PMA from WT.A total of 203 patients with histologically confirmed PMA or WT who underwent parotidectomy from two centers were enrolled. Clinical factors, ultrasound (US) features, and radiomic features were extracted to develop three types of machine learning model: clinical models, US models, and USEML models. The diagnostic performance of the USEML model, as well as that of physicians based on experience, was evaluated and validated using receiver operating characteristic (ROC) curves in internal and external validation cohorts. DeLong’s test was used for comparisons of AUCs. SHAP values were also utilized to explain the classification model.The USEML model achieved the highest AUC of 0.891 (95% CI, 0.774–0.961), surpassing the AUCs of both the US (0.847; 95% CI, 0.720–0.932) and clinical (0.814; 95% CI, 0.682–0.908) models. The USEML model also outperformed physicians in both internal and external validation datasets (both p < 0.05). The sensitivity, specificity, negative predictive value, and positive predictive value of the USEML model and physician experience were 89.3%/75.0%, 87.5%/54.2%, 87.5%/65.6%, and 89.3%/65.0%, respectively.The USEML model, incorporating clinical factors, ultrasound factors, and radiomic features, demonstrated efficient performance in distinguishing PMA from WT in the parotid gland.This study developed a machine learning model for preoperative diagnosis of pleomorphic adenoma and Warthin tumor in the parotid gland based on clinical, ultrasound, and radiomic features. Furthermore, it outperformed physicians in an external validation dataset, indicating its potential for clinical application.• Differentiating pleomorphic adenoma (PMA) and Warthin tumor (WT) affects management decisions and is currently done by invasive biopsy.• Integration of US-radiomic, clinical, and ultrasound findings in a machine learning model results in improved diagnostic accuracy.• The ultrasound-based ensemble machine learning (USEML) model consistently outperforms physicians, suggesting its potential applicability in clinical settings.Clinical relevance statement: The preoperative classification of pleomorphic adenomas (PMA) and Warthin tumors (WT) in the parotid gland plays an essential role in determining therapeutic strategies. This study aims to develop and validate an ultrasound-based ensemble machine learning (USEML) model, employing nonradiative and noninvasive features to differentiate PMA from WT.A total of 203 patients with histologically confirmed PMA or WT who underwent parotidectomy from two centers were enrolled. Clinical factors, ultrasound (US) features, and radiomic features were extracted to develop three types of machine learning model: clinical models, US models, and USEML models. The diagnostic performance of the USEML model, as well as that of physicians based on experience, was evaluated and validated using receiver operating characteristic (ROC) curves in internal and external validation cohorts. DeLong’s test was used for comparisons of AUCs. SHAP values were also utilized to explain the classification model.The USEML model achieved the highest AUC of 0.891 (95% CI, 0.774–0.961), surpassing the AUCs of both the US (0.847; 95% CI, 0.720–0.932) and clinical (0.814; 95% CI, 0.682–0.908) models. The USEML model also outperformed physicians in both internal and external validation datasets (both p < 0.05). The sensitivity, specificity, negative predictive value, and positive predictive value of the USEML model and physician experience were 89.3%/75.0%, 87.5%/54.2%, 87.5%/65.6%, and 89.3%/65.0%, respectively.The USEML model, incorporating clinical factors, ultrasound factors, and radiomic features, demonstrated efficient performance in distinguishing PMA from WT in the parotid gland.This study developed a machine learning model for preoperative diagnosis of pleomorphic adenoma and Warthin tumor in the parotid gland based on clinical, ultrasound, and radiomic features. Furthermore, it outperformed physicians in an external validation dataset, indicating its potential for clinical application.• Differentiating pleomorphic adenoma (PMA) and Warthin tumor (WT) affects management decisions and is currently done by invasive biopsy.• Integration of US-radiomic, clinical, and ultrasound findings in a machine learning model results in improved diagnostic accuracy.• The ultrasound-based ensemble machine learning (USEML) model consistently outperforms physicians, suggesting its potential applicability in clinical settings.Key Points: The preoperative classification of pleomorphic adenomas (PMA) and Warthin tumors (WT) in the parotid gland plays an essential role in determining therapeutic strategies. This study aims to develop and validate an ultrasound-based ensemble machine learning (USEML) model, employing nonradiative and noninvasive features to differentiate PMA from WT.A total of 203 patients with histologically confirmed PMA or WT who underwent parotidectomy from two centers were enrolled. Clinical factors, ultrasound (US) features, and radiomic features were extracted to develop three types of machine learning model: clinical models, US models, and USEML models. The diagnostic performance of the USEML model, as well as that of physicians based on experience, was evaluated and validated using receiver operating characteristic (ROC) curves in internal and external validation cohorts. DeLong’s test was used for comparisons of AUCs. SHAP values were also utilized to explain the classification model.The USEML model achieved the highest AUC of 0.891 (95% CI, 0.774–0.961), surpassing the AUCs of both the US (0.847; 95% CI, 0.720–0.932) and clinical (0.814; 95% CI, 0.682–0.908) models. The USEML model also outperformed physicians in both internal and external validation datasets (both p < 0.05). The sensitivity, specificity, negative predictive value, and positive predictive value of the USEML model and physician experience were 89.3%/75.0%, 87.5%/54.2%, 87.5%/65.6%, and 89.3%/65.0%, respectively.The USEML model, incorporating clinical factors, ultrasound factors, and radiomic features, demonstrated efficient performance in distinguishing PMA from WT in the parotid gland.This study developed a machine learning model for preoperative diagnosis of pleomorphic adenoma and Warthin tumor in the parotid gland based on clinical, ultrasound, and radiomic features. Furthermore, it outperformed physicians in an external validation dataset, indicating its potential for clinical application.• Differentiating pleomorphic adenoma (PMA) and Warthin tumor (WT) affects management decisions and is currently done by invasive biopsy.• Integration of US-radiomic, clinical, and ultrasound findings in a machine learning model results in improved diagnostic accuracy.• The ultrasound-based ensemble machine learning (USEML) model consistently outperforms physicians, suggesting its potential applicability in clinical settings. [ABSTRACT FROM AUTHOR]