RESUMEN

The air pollution levels from polychlorodibenzo-p-dioxins/polychlorodibenzofurans (PCDD/Fs) and dioxin-like polychlorobiphenyls (dl-PCBs) in three residential areas located north, west, and south of the Da Nang airport were determined by using passive air samplers containing polyurethane foam (PUF) discs with 3-month sampling intervals from 2017 to 2020. The total toxic equivalents (∑TEQs) of the PCDD/Fs and dl-PCBs, using WHO2005-TEFs, were highest north of the airport (134 to 10610 fg WHO-TEQ/PUF day, with an average of 1108 fg WHO-TEQ/PUF day). The ∑TEQs were lower west of the airport, between 159 and 381 fg WHO-TEQ/PUF day and averaged 230 fg WHO-TEQ/PUF day. The lowest ∑TEQs occurred south of the airport, with ranges of 76 and 331 fg WHO-TEQ/PUF day and an average of 152 fg WHO-TEQ/PUF day. Construction activities, including excavation and transportation of dioxin-contaminated soil north of the airport, have increased airborne PCDD/F and dl-PCB contamination and health risks. The average daily doses of PCDD/Fs and dl-PCBs through inhalation (ADDA) for residents located north of the airport were the highest (10.9 to 3434 fg WHO-TEQ/kg BW/day and average: 597 fg WHO-TEQ/kg BW/day). Residents located west of the airport faced lower health risks (13-123 fg WHO-TEQ/kg BW/day and average: 39 fg WHO-TEQ/kg BW/day). Residents south of the airport were exposed to a minimum of 6.2-107 fg WHO-TEQ/kg BW/day, with an average of 28 fg WHO-TEQ/kg BW/day. The maximum and average ADDA values for residents north of the airport exceeded 10% of the tolerable daily intake (TDI) recommended by the WHO (100-400 fg WHO-TEQ/kg BW/day). In comparison, all the ADDA values for residents located west and south of the airport were less than and within 10% of the TDI.

RESUMEN

Recent artificial intelligence (AI) advancements in cardiovascular medicine offer potential enhancements in diagnosis, prediction, treatment, and outcomes. This article aims to provide a basic understanding of AI enabled ECG technology. Specific conditions and findings will be discussed, followed by reviewing associated terminology and methodology. In the appendix, definitions of AUC versus accuracy are explained. The application of deep learning models enables detecting diseases from normal electrocardiograms at accuracy not previously achieved by technology or human experts. Results with AI enabled ECG are encouraging as they considerably exceeded current screening models for specific conditions (i.e., atrial fibrillation, left ventricular dysfunction, aortic stenosis, and hypertrophic cardiomyopathy). This could potentially lead to a revitalization of the utilization of the ECG in the insurance domain. While we are embracing the findings with this rapidly evolving technology, but cautious optimism is still necessary at this point.

Asunto(s)

RESUMEN

Advanced bioinformatics analysis, such as systems biology (SysBio) and artificial intelligence (AI) approaches, including machine learning (ML) and deep learning (DL), is increasingly present in stem cell (SC) research. An approximate timeline on these developments and their global impact is still lacking. We conducted a scoping review on the contribution of SysBio and AI analysis to SC research and therapy development based on literature published in PubMed between 2000 and 2024. We identified an 8-10-fold increase in research output related to all three search terms between 2000 and 2021, with a 10-fold increase in AI-related production since 2010. Use of SysBio and AI still predominates in preclinical basic research with increasing use in clinically oriented translational medicine since 2010. SysBio- and AI-related research was found all over the globe, with SysBio output led by the United States (US, n=1487), United Kingdom (UK, n=1094), Germany (n=355), The Netherlands (n=339), Russia (n=215), and France (n=149), while for AI-related research the US (n=853) and UK (n=258) take a strong lead, followed by Switzerland (n=69), The Netherlands (n=37), and Germany (n=19). The US and UK are most active in SCs publications related to AI/ML and AI/DL. The prominent use of SysBio in ESC research was recently overtaken by prominent use of AI in iPSC and MSC research. This study reveals the global evolution and growing intersection between AI, SysBio, and SC research over the past two decades, with substantial growth in all three fields and exponential increases in AI-related research in the past decade.

RESUMEN

Brocchia cinerea is a North African plant belonging to the Asteraceae family, widely utilized in Algerian folk medicine to treat a variety of illnesses. These therapeutic virtues are mainly due to the plant essential oil. The chemical components of this oil were identified using GC-MS, and the variability in these components' levels was examined in nine samples that were taken at different times from two locations in Algeria's northern Sahara. The contents of the essential oil were found to consist of eight components, varying in concentrations: beta-thujone (46.80%), 1-Methyl-2-(1' methylethenyl) -3'- ethenylcyclopropylmethanol (14.59%), 1,8-Cineole (12.63%), limonen-10-ol (9.47%), 1(7),3,8 o Menthatriene (3.45%), and (-)-Camphor (2.11%). Toxicity studies were conducted in order to assess the safety of the essential oil, namely: LD50 estimation and biochemical blood parameters evaluation. The results showed an LD 50 of 507.5mg/kg close to the LD50 of Beta-thujone (442mg/kg) : the main component of the essential oil, making it accountable for the major toxicity. The apparition of seizures as toxic manifestations for higher concentrations confirmed that. The essential oil of Brocchia was noted to be classified as slightly, weakly toxic, and the Beta-thujone contents showed to be within the regulatory accepted values, which makes the use of Brocchia safe within the indicated standards.

RESUMEN

Subarachnoid hemorrhage (SAH) is a severe neurological event lacking of effective therapy. Early brain injury (EBI) and delayed neurological dysfunction are important cause in the poor prognosis of patients with SAH. Nucleotide-binding oligomerization domain (NOD)-like receptor pyrin domain containing 3 (NLRP3) inflammasome activation has been implicated in many inflammatory lesion pathogeneses including SAH. Dl-3-n-butylphthalide (NBP) has been reported to possess substantial anti-inflammatory properties, which is beneficial for various neurodegenerative diseases. However, the effect and molecular mechanisms of NBP on SAH have not been clearly identified. We designed this study to investigate the effect of NBP against EBI and delayed neurological dysfunction after SAH and to reveal the possible underlying mechanism. The adult mice were subjected to endovascular perforation SAH model or sham operation. Mice were randomized to sham group, SAH group, or SAH+NBP group. The EBI (short-term study) was studied at 48h post-SAH and delayed neurological dysfunction (long-term study) at 21 days post-SAH. The results suggested that NBP evidently alleviated the EBI in mice at 48h post-SAH, as shown by elevating neurological score, reducing brain edema, blood-brain barrier disruption, neuronal loss, and astrocyte aggregation, as well as ameliorating cerebral vasospasm. Moreover, NBP was able to improve long-term neurobehavioral functions and decrease neuronal apoptosis at 21 days after SAH. Significantly, NBP treatment also inhibited the expressions of NLRP3, ASC, caspase-1, cleaved-caspase-1, IL-1ß, IL-18, GSDMD and GSDMD-N in both EBI and delayed neurological dysfunction indued by SAH. Our findings suggested that NBP treatment exerts a profound neuroprotective effect against early brain injury and delayed neurological dysfunction induced by SAH, at least partially through regulating NLRP3 inflammasome signaling pathway and its related inflammation and pyroptosis.

RESUMEN

Passive air samplers were used to monitor polychlorodibenzo-p-dioxins and dibenzofurans (PCDD/Fs) and dioxin-like polychlorobiphenyls (dl-PCBs) between 2020 and 2022 in four residential areas around Bien Hoa hot spot (BHS) including Trung Dung (TD), Tan Phong (TP), Quang Vinh (QV), and Buu Long (BL). The total toxic equivalents of PCDD/Fs and dl-PCBs (∑TEQs) were highest in the TD area, from 284 to 642 fg TEQ/PUF day. Next was the QV area, where ∑TEQs ranged from 229 to 569 fg TEQ/PUF day. Then, ∑TEQs varied from 205 to 503 fg TEQ/PUF day in the TP area. The lowest ∑TEQs were between 179 and 385 fg TEQ/PUF day in the BL area. The temporal, spatial, and seasonal variations in concentrations of PCDD/Fs and dl-PCBs were related to the prevailing wind direction and the distance from each area to the dioxin hot spot. The average ∑TEQs for all four areas surrounding BHS in the dry season (423 fg TEQ/PUF day) were 1.4 times higher than in the rainy season (303 fg TEQ/PUF day). Health risk assessments from airborne dioxin exposure were estimated using the average daily doses through inhalation (ADDI). The ADDI for residents surrounding BHS ranged from 14.6 to 208 fg TEQ/kg BW/day. The ADDI values by areas were as follows: 23.2-208 fg TEQ/kg BW/day in the TD, 18.7-184 fg TEQ/kg BW/day in the QV, 16.7-163 fg TEQ/kg BW/day in the TP, and 14.6-125 fg TEQ/kg BW/day in the BL. These ADDI values remained within and below the 10% threshold of the WHO-recommended tolerable daily intake (100-400 fg TEQ/kg BW/day). It is necessary to control the excavation activities inside the BHS and cover the temporary storage sites of dioxin-contaminated materials to minimize the emissions of PCDD/Fs and dl-PCB into the ambient air.

RESUMEN

Endogenous enzymes play a crucial role in determining fish product aroma. However, the attached microorganisms can promote enzyme production, making it challenging to identify specific aromatic compounds resulting from endogenous enzymes. Thus, we investigated the aroma transformation of Japanese sea bass through enzymatic incubation by controlling attached microorganisms during the lag phase. Our results demonstrate that enzymatic incubation significantly enhances grassy and sweet notes while reducing fishy odors. These changes in aroma are associated with increased levels of 10 volatile compounds and decreased levels of 3 volatile compounds. Among them, previous studies have reported enzyme reaction pathways for octanal, 1-nonanal, vanillin, indole, linalool, geraniol, citral, and 6-methyl-5-hepten-2-one; however, the enzymatic reaction pathways for germacrene D, beta-caryophyllene, pristane, 1-tetradecene and trans-beta-ocimene remain unclear. These findings provide novel insights for further study to elucidate the impact of endogenous enzymes on fish product aromas.

RESUMEN

The application of artificial intelligence (AI) in electrocardiography is revolutionizing cardiology and providing essential insights into the consequences of the COVID-19 pandemic. This comprehensive review explores AI-enhanced ECG (AI-ECG) applications in risk prediction and diagnosis of heart diseases, with a dedicated chapter on COVID-19-related complications. Introductory concepts on AI and machine learning (ML) are explained to provide a foundational understanding for those seeking knowledge, supported by examples from the literature and current practices. We analyze AI and ML methods for arrhythmias, heart failure, pulmonary hypertension, mortality prediction, cardiomyopathy, mitral regurgitation, hypertension, pulmonary embolism, and myocardial infarction, comparing their effectiveness from both medical and AI perspectives. Special emphasis is placed on AI applications in COVID-19 and cardiology, including detailed comparisons of different methods, identifying the most suitable AI approaches for specific medical applications and analyzing their strengths, weaknesses, accuracy, clinical relevance, and key findings. Additionally, we explore AI's role in the emerging field of cardio-oncology, particularly in managing chemotherapy-induced cardiotoxicity and detecting cardiac masses. This comprehensive review serves as both an insightful guide and a call to action for further research and collaboration in the integration of AI in cardiology, aiming to enhance precision medicine and optimize clinical decision-making.

RESUMEN

BACKGROUND: The pretherapeutic differentiation of subtypes of primary intracranial germ cell tumours (iGCTs), including germinomas (GEs) and nongerminomatous germ cell tumours (NGGCTs), is essential for clinical practice because of distinct treatment strategies and prognostic profiles of these diseases. This study aimed to develop a deep learning model, iGNet, to assist in the differentiation and prognostication of iGCT subtypes by employing pretherapeutic MR T2-weighted imaging. METHODS: The iGNet model, which is based on the nnUNet architecture, was developed using a retrospective dataset of 280 pathologically confirmed iGCT patients. The training dataset included 83 GEs and 117 NGGCTs, while the retrospective internal test dataset included 31 GEs and 49 NGGCTs. The model's diagnostic performance was then assessed with the area under the receiver operating characteristic curve (AUC) in a prospective internal dataset (n = 22) and two external datasets (n = 22 and 20). Next, we compared the diagnostic performance of six neuroradiologists with or without the assistance of iGNet. Finally, the predictive ability of the output of iGNet for progression-free and overall survival was assessed and compared to that of the pathological diagnosis. RESULTS: iGNet achieved high diagnostic performance, with AUCs between 0.869 and 0.950 across the four test datasets. With the assistance of iGNet, the six neuroradiologists' diagnostic AUCs (averages of the four test datasets) increased by 9.22% to 17.90%. There was no significant difference between the output of iGNet and the results of pathological diagnosis in predicting progression-free and overall survival (P = .889). CONCLUSIONS: By leveraging pretherapeutic MR imaging data, iGNet accurately differentiates iGCT subtypes, facilitating prognostic evaluation and increasing the potential for tailored treatment.

Asunto(s)

RESUMEN

Deep Learning (DL) drives academics to create models for cancer diagnosis using medical image processing because of its innate ability to recognize difficult-to-detect patterns in complex, noisy, and massive data. The use of deep learning algorithms for real-time cancer diagnosis is explored in depth in this work. Real-time medical diagnosis determines the illness or condition that accounts for a patient's symptoms and outward physical manifestations within a predetermined time frame. With a waiting period of anywhere between 5 days and 30 days, there are currently several ways, including screening tests, biopsies, and other prospective methods, that can assist in discovering a problem, particularly cancer. This article conducts a thorough literature review to understand how DL affects the length of this waiting period. In addition, the accuracy and turnaround time of different imaging modalities is evaluated with DL-based cancer diagnosis. Convolutional neural networks are critical for real-time cancer diagnosis, with models achieving up to 99.3% accuracy. The effectiveness and cost of the infrastructure required for real-time image-based medical diagnostics are evaluated. According to the report, generalization problems, data variability, and explainable DL are some of the most significant barriers to using DL in clinical trials. Making DL applicable for cancer diagnosis will be made possible by explainable DL.

RESUMEN

Sarcasm in Sentiment Analysis (SA) is important due to the sense of sarcasm in sentences that differs from their literal meaning. Analysis of Arabic sarcasm still has many challenges like implicit indirect idioms to express the opinion, and lack of Arabic sarcasm corpus. In this paper, we proposed a new detecting model for sarcasm in Arabic tweets called the ArSa-Tweet model. It is based on implementing and developing Deep Learning (DL) models to classify tweets as sarcastic or not. The development of our proposed model consists of adding main improvements by applying robust preprocessing steps before feeding the data to the adapted DL models. The adapted DL models are LSTM, Multi-headed CNN-LSTM-GRU, BERT, AraBert-V01, and AraBert-V02. In addition, we proposed ArSa-data as a golden corpus that consists of Arabic tweets. A comparative process shows that our proposed ArSa-Tweet method has the most impact accuracy rate based on deploying the AraBert-V02 model, which obtains the best performance results in all accuracy metrics when compared with other methods.

RESUMEN

KIR3DL3*0070105, KIR3DL3*0130202, KIR3DL1*0080104 and KIR3DL1*0010121, identified by next generation sequencing in individuals from India.

Asunto(s)

RESUMEN

The novel KIR2DL1*00308 allele differs from the closest allele KIR2DL1*00302 by a single sense mutation.

Asunto(s)

RESUMEN

Many vaccine design programs have been developed, including our own machine learning approaches Vaxign-ML and Vaxign-DL. Using deep learning techniques, Vaxign-DL predicts bacterial protective antigens by calculating 509 biological and biomedical features from protein sequences. In this study, we first used the protein folding ESM program to calculate a set of 1,280 features from individual protein sequences, and then utilized the new set of features separately or in combination with the traditional set of 509 features to predict protective antigens. Our result showed that the usage of ESM-derived features alone was able to accurately predict vaccine antigens with a performance similar to the orginal Vaxign-DL prediction method, and the usage of the combined ESM-derived and orginal Vaxign-DL features significantly improved the prediction performance according to a set of seven scores including specificity, sensitivity, and AUROC. To further evaluate the updated methods, we conducted a Leave-One-Pathogen-Out Validation (LOPOV) study, and found that the usage of ESM-derived features significantly improved the the prediction of vaccine antigens from 10 bacterial pathogens. This research is the first reported study demonstrating the added value of protein folding features for vaccine antigen prediction.

RESUMEN

The ubiquitous presence of electronic devices demands robust hardware security mechanisms to safeguard sensitive information from threats. This paper presents a physical unclonable function (PUF) circuit based on magnetoresistive random access memory (MRAM). The circuit utilizes inherent characteristics arising from fabrication variations, specifically magnetic tunnel junction (MTJ) cell resistance, to produce corresponding outputs for applied challenges. In contrast to Arbiter PUF, the proposed effectively satisfies the strict avalanche criterion (SAC). Additionally, the grid-like structure of the proposed circuit preserves its resistance against machine learning-based modeling attacks. Various machine learning (ML) attacks employing multilayer perceptron (MLP), linear regression (LR), and support vector machine (SVM) networks are simulated for two-array and four-array architectures. The MLP-attack prediction accuracy was 53.61% for a two-array circuit and 49.87% for a four-array circuit, showcasing robust performance even under the worst-case process variations. In addition, deep learning-based modeling attacks in considerable high dimensions utilizing multiple networks such as convolutional neural network (CNN), recurrent neural network (RNN), MLP, and Larq are used with the accuracy of 50.31%, 50.25%, 50.31%, and 50.31%, respectively. The efficiency of the proposed circuit at the layout level is also investigated for simplified two-array architecture. The simulation results indicate that the proposed circuit offers intra and inter-hamming distance (HD) with a mean of 0.98% and 49.96%, respectively, and a mean diffuseness of 49.09%.

RESUMEN

Background: Diabetic kidney disease is a major contributor to end stage renal disease. A change in kidney oxygen homeostasis leading to decreased tissue oxygen tension is an important factor initiating alterations in kidney function in diabetes. However, the mechanism contributing to changed oxygen homeostasis is still unclear. Hyperglycemia-induced production of reactive oxygen species and an altered response to them have previously been demonstrated. In the present study, chronic treatment with DL-sulforaphane to induce nuclear factor erythroid 2-related factor 2 (Nrf2) expression, a master transcriptional regulator binding to antioxidant response elements inducing increased protection against reactive oxygen species, is studied. Methods: Sprague-Dawley rats were made diabetic using streptozotocin and either left untreated or received daily subcutaneous injections of DL-sulforaphane for 4 weeks. Age-matched non-diabetic rats served as controls. After 4 weeks of treatment, rats were anesthetized using thiobutabarbital, and kidney functions were studied in terms of glomerular filtration rate (GFR), renal blood flow (RBF), sodium transport, kidney oxygen consumption, and kidney oxygen tension. Mitochondria was isolated from kidney cortical tissue and investigated using high-resolution respirometry. Results: GFR was increased in diabetics but not RBF resulting in increased filtration fraction in diabetics. DL-sulforaphane treatment did not affect RBF and GFR in controls but decreased the same parameters in diabetics. Increased GFR resulted in increased sodium transport and oxygen consumption, hence decreased efficiency in diabetics compared to controls. Increased oxygen consumption in diabetics resulted in decreased cortical tissue oxygen tension. DL-sulforaphane treatment decreased oxygen consumption in diabetics, whereas transport efficiency was not significantly affected. DL-sulforaphane treatment increased cortical pO2 in diabetics. Conclusions: DL-sulforaphane treatment affects renal hemodynamics, improving cortical oxygen tension but not mitochondrial efficiency.

Asunto(s)

RESUMEN

The detection of the cognitive tasks performed by a subject during data acquisition of a neuroimaging method has a wide range of applications: functioning of brain-computer interface (BCI), detection of neuronal disorders, neurorehabilitation for disabled patients, and many others. Recent studies show that the combination or fusion of electroencephalography (EEG) and functional near-infrared spectroscopy (fNIRS) demonstrates improved classification and detection performance compared to sole-EEG and sole-fNIRS. Deep learning (DL) networks are suitable for the classification of large volume time-series data like EEG and fNIRS. This study performs the decision fusion of EEG and fNIRS. The classification of EEG, fNIRS, and decision-fused EEG-fNIRSinto cognitive task labels is performed by DL networks. Two different open-source datasets of simultaneously recorded EEG and fNIRS are examined in this study. Dataset 01 is comprised of 26 subjects performing 3 cognitive tasks: n-back, discrimination or selection response (DSR), and word generation (WG). After data acquisition, fNIRS is converted to oxygenated hemoglobin (HbO2) and deoxygenated hemoglobin (HbR) in Dataset 01. Dataset 02 is comprised of 29 subjects who performed 2 tasks: motor imagery and mental arithmetic. The classification procedure of EEG and fNIRS (or HbO2, HbR) are carried out by 7 DL classifiers: convolutional neural network (CNN), long short-term memory network (LSTM), gated recurrent unit (GRU), CNN-LSTM, CNN-GRU, LSTM-GRU, and CNN-LSTM-GRU. After the classification of single modalities, their prediction scores or decisions are combined to obtain the decision-fused modality. The classification performance is measured by overall accuracy and area under the ROC curve (AUC). The highest accuracy and AUC recorded in Dataset 01 are 96% and 100% respectively; both by the decision fusion modality using CNN-LSTM-GRU. For Dataset 02, the highest accuracy and AUC are 82.76% and 90.44% respectively; both by the decision fusion modality using CNN-LSTM. The experimental result shows that decision-fused EEG-HbO2-HbR and EEG-fNIRSdeliver higher performances compared to their constituent unimodalities in most cases. For DL classifiers, CNN-LSTM-GRU in Dataset 01 and CNN-LSTM in Dataset 02 yield the highest performance.

RESUMEN

Background: Lumbar spine disorders are one of the common causes of low back pain (LBP). Objective and reliable measurement of anatomical parameters of the lumbar spine is essential in the clinical diagnosis and evaluation of lumbar disorders. However, manual measurements are time-consuming and laborious, with poor consistency and repeatability. Here, we aim to develop and evaluate an automatic measurement model for measuring the anatomical parameters of the vertebral body and intervertebral disc based on lateral lumbar radiographs and deep learning (DL). Methods: A model based on DL was developed with a dataset consisting of 1,318 lateral lumbar radiographs for the prediction of anatomical parameters, including vertebral body heights (VBH), intervertebral disc heights (IDH), and intervertebral disc angles (IDA). The mean of the values obtained by 3 radiologists was used as a reference standard. Statistical analysis was performed in terms of standard deviation (SD), mean absolute error (MAE), Percentage of correct keypoints (PCK), intraclass correlation coefficient (ICC), regression analysis, and Bland-Altman plot to evaluate the performance of the model compared with the reference standard. Results: The percentage of intra-observer landmark distance within the 3 mm threshold was 96%. The percentage of inter-observer landmark distance within the 3 mm threshold was 94% (R1 and R2), 92% (R1 and R3), and 93% (R2 and R3), respectively. The PCK of the model within the 3 mm distance threshold was 94-99%. The model-predicted values were 30.22±3.01 mm, 10.40±3.91 mm, and 10.63°±4.74° for VBH, IDH, and IDA, respectively. There were good correlation and consistency in anatomical parameters of the lumbar vertebral body and disc between the model and the reference standard in most cases (R2=0.89-0.95, ICC =0.93-0.98, MAE =0.61-1.15, and SD =0.89-1.64). Conclusions: The newly proposed model based on a DL algorithm can accurately measure various anatomical parameters on lateral lumbar radiographs. This could provide an accurate and efficient measurement tool for the quantitative evaluation of spinal disorders.

RESUMEN

Background: Most primary bone tumors are often found in the bone around the knee joint. However, the detection of primary bone tumors on radiographs can be challenging for the inexperienced or junior radiologist. This study aimed to develop a deep learning (DL) model for the detection of primary bone tumors around the knee joint on radiographs. Methods: From four tertiary referral centers, we recruited 687 patients diagnosed with bone tumors (including osteosarcoma, chondrosarcoma, giant cell tumor of bone, bone cyst, enchondroma, fibrous dysplasia, etc.; 417 males, 270 females; mean age 22.8±13.2 years) by postoperative pathology or clinical imaging/follow-up, and 1,988 participants with normal bone radiographs (1,152 males, 836 females; mean age 27.9±12.2 years). The dataset was split into a training set for model development, an internal independent and an external test set for model validation. The trained model located bone tumor lesions and then detected tumor patients. Receiver operating characteristic curves and Cohen's kappa coefficient were used for evaluating detection performance. We compared the model's detection performance with that of two junior radiologists in the internal test set using permutation tests. Results: The DL model correctly localized 94.5% and 92.9% bone tumors on radiographs in the internal and external test set, respectively. An accuracy of 0.964/0.920, and an area under the receiver operating characteristic curve (AUC) of 0.981/0.990 in DL detection of bone tumor patients were for the internal and external test set, respectively. Cohen's kappa coefficient of the model in the internal test set was significantly higher than that of the two junior radiologists with 4 and 3 years of experience in musculoskeletal radiology (Model vs. Reader A, 0.927 vs. 0.777, P<0.001; Model vs. Reader B, 0.927 vs. 0.841, P=0.033). Conclusions: The DL model achieved good performance in detecting primary bone tumors around the knee joint. This model had better performance than those of junior radiologists, indicating the potential for the detection of bone tumors on radiographs.

RESUMEN

Background: The incidence rate of thyroid nodules has reached 65%, but only 5-15% of these modules are malignant. Therefore, accurately determining the benign and malignant nature of thyroid nodules can prevent unnecessary treatment. We aimed to develop a deep-learning (DL) radiomics model based on ultrasound (US), explore its diagnostic efficacy for benign and malignant thyroid nodules, and verify whether it improved the diagnostic level of physicians. Methods: We retrospectively included 1,076 thyroid nodules from 817 patients at three institutions. The radiomics and DL features of the US images were extracted and used to construct radiomics signature (Rad_sig) and deep-learning signature (DL_sig). A Pearson correlation analysis and least absolute shrinkage and selection operator (LASSO) regression analysis were used for feature selection. Clinical US semantic signature (C_US_sig) was constructed based on clinical information and US semantic features. Next, a combined model was constructed based on the above three signatures in the form of a nomogram. The model was constructed using a development set (institution 1: 719 nodules), and the model was evaluated using two external validation sets (institution 2: 74 nodules, and institution 3: 283 nodules). The performance of the model was assessed using decision curve analysis (DCA) and calibration curves. Furthermore, the C_US_sigs of junior physicians, senior physicians, and expers were constructed. The DL radiomics model was used to assist the physicians with different levels of experience in the interpretation of thyroid nodules. Results: In the development and validation sets, the combined model showed the highest performance, with areas under the curve (AUCs) of 0.947, 0.917, and 0.929, respectively. The DCA results showed that the comprehensive nomogram had the best clinical utility. The calibration curves indicated good calibration for all models. The AUCs for distinguishing between benign and malignant thyroid nodules by junior physicians, senior physicians, and experts were 0.714-0.752, 0.740-0.824, and 0.891-0.908, respectively; however, with the assistance of DL radiomics, the AUCs reached 0.858-0.923, 0.888-0.944, and 0.912-0.919, respectively. Conclusions: The nomogram based on DL radiomics had high diagnostic efficacy for thyroid nodules, and DL radiomics could assist physicians with different levels of experience to improve their diagnostic level.