Receiver Operating Characteristic Area Under The Curve Research Articles

Inpatient Measurements of Urine Metanephrines are Indistinguishable from Pheochromocytoma: Retrospective Cohort Study

BackgroundPheochromocytoma is a rare cause of acute cardiovascular disease; however, any severe illness may have high catecholamines, simulating pheochromocytoma. We determined the spectrum of urine metanephrines from inpatient and outpatient collections without pheochromocytoma, compared with confirmed pheochromocytoma patients. MethodsRetrospective analysis using centralized laboratory data serving all outpatients and hospitals in southern Alberta. The analysis comprised 24-hour urine normetanephrine and metanephrine (UNM-UMN) results collected from hospital inpatients, community outpatients, and patients from a comprehensive provincial pheochromocytoma registry. ResultsThere were 974 unique inpatients (including 132 from intensive care), 6802 outpatients, and 58 pheochromocytoma patients. Among outpatient, general ward, and intensive care unit (ICU) patients, 18.7%, 34.4%, and 67.4% of results, respectively, were supranormal. Although pheochromocytoma patients had higher median UNM-UMN vs inpatients, there was substantial overlap. Receiver operating characteristic (ROC) analysis showed area under the curve (AUC) of 0.64-0.91 to detect true pheochromocytoma (P < .0001), with progressively poorer discrimination among hospitalized and ICU-dependent patients. A 24-hour urine normetanephrine >6.95 nmol/d had 98% specificity for pheochromocytoma when inpatient general ward samples were included, but only 46% sensitivity and 13% positive predictive value for pheochromocytoma. Considering ICU collections, 98% specificity required results more than fivefold above the upper reference limit and still had poor positive predictive value. A model combining both UNM and UMN results as a cross-product marginally improved the ROC AUC, but improved sensitivity in outpatients and ward patients but not ICU patients. ConclusionThere is a high degree of overlap in UNM-UMN between hospitalized patients and pheochromocytoma; high test specificity is not achieved in this population unless >3-5 times the upper reference limit.

The value of human epididymis 4, D-dimer, and fibrinogen compared with CA125 alone in triaging women presenting with pelvic masses: a retrospective cohort study.

CA125, the biomarker in common clinical use for ovarian cancer, is limited by low sensitivity for early disease and high false positives. The aim of this study was to evaluate several candidate biomarkers, alone or in combination, compared with CA125 in the prediction of malignant/borderline vs benign tumor status in premenopausal and postmenopausal women with pelvic masses. This was a retrospective observational cohort study set in St James's Hospital, a tertiary referral center for gynecological malignancy in Dublin, Ireland. Women undergoing surgery for pelvic masses between 2012 and 2018 were included. Preoperative human epididymis protein 4 (HE4), the Risk of Ovarian Malignancy Algorithm, the Risk of Malignancy Index I and II, D-dimer, and fibrinogen were assessed. Logistic regression models were fitted for each biomarker alone and in combination. Receiver operating characteristics-area under the curve (ROC-AUC) and partial AUCs in the 90%-100% specificity range were determined. In all, 89 premenopausal and 185 postmenopausal women were included. In premenopausal women, no biomarker(s) outperformed CA125 (AUC 0.73; 95% CI 0.63-0.84). In postmenopausal women, HE4 had a partial AUC (pAUC) of 0.71 (95% CI 0.64-0.79) compared with 0.57 (95% CI 0.51-0.69) for CA125 (p=0.009). HE4 + D-dimer had an improved pAUC of 0.74 (95% CI 0.68-0.81, p<0.001) and HE4 + D-dimer + fibrinogen had a pAUC of 0.75 (95% CI 0.68-0.82). A novel biomarker panel of HE4±D-dimer±fibrinogen outperformed CA125 alone as a high-specificity biomarker in postmenopausal women and could aid in the preoperative triaging of pelvic masses. No biomarker(s) outperformed CA125 in premenopausal women.

Identification of SARS-CoV-2 viral entry inhibitors using machine learning and cell-based pseudotyped particle assay

In response to the pandemic caused by SARS-CoV-2, we constructed a hybrid support vector machine (SVM) classification model using a set of publicly posted SARS-CoV-2 pseudotyped particle (PP) entry assay repurposing screen data to identify novel potent compounds as a starting point for drug development to treat COVID-19 patients. Two different molecular descriptor systems, atom typing descriptors and 3D fingerprints (FPs), were employed to construct the SVM classification models. Both models achieved reasonable performance, with the area under the curve of receiver operating characteristic (AUC-ROC) of 0.84 and 0.82, respectively. The consensus prediction outperformed the two individual models with significantly improved AUC-ROC of 0.91, where the compounds with inconsistent classifications were excluded. The consensus model was then used to screen the 173,898 compounds in the NCATS annotated and diverse chemical libraries. Of the 255 compounds selected for experimental confirmation, 116 compounds exhibited inhibitory activities in the SARS-CoV-2 PP entry assay with IC50 values ranged between 0.17 µM and 62.2 µM, representing an enrichment factor of 3.2. These 116 active compounds with diverse and novel structures could potentially serve as starting points for chemistry optimization for COVID-19 drug discovery.

Identification of a tumor microenvironment-related gene signature to improve the prediction of cervical cancer prognosis

BackgroundPrevious studies have found that the microenvironment of cervical cancer (CESC) affects the progression and treatment of this disease. Thus, we constructed a multigene model to assess the survival of patients with cervical cancer.MethodsWe scored 307 CESC samples from The Cancer Genome Atlas (TCGA) and divided them into high and low matrix and immune scores using the ESTIMATE algorithm for differential gene analysis. Cervical cancer patients were randomly divided into a training group, testing group and combined group. The multigene signature prognostic model was constructed by Cox analyses. Multivariate Cox analysis was applied to evaluate the significance of the multigene signature for cervical cancer prognosis. Prognosis was assessed by Kaplan–Meier curves comparing the different groups, and the accuracy of the prognostic model was analyzed by receiver operating characteristic-area under the curve (ROC-AUC) analysis and calibration curve. The Tumor Immune Estimation Resource (TIMER) database was used to analyze the relationship between the multigene signature and immune cell infiltration.ResultsWe obtained 420 differentially expressed genes in the tumor microenvironment from 307 patients with cervical cancer. A three-gene signature (SLAMF1, CD27, SELL) model related to the tumor microenvironment was constructed to assess patient survival. Kaplan–Meier analysis showed that patients with high risk scores had a poor prognosis. The ROC-AUC value indicated that the model was an accurate predictor of cervical cancer prognosis. Multivariate cox analysis showed the three-gene signature to be an independent risk factor for the prognosis of cervical cancer. A nomogram combining the three-gene signature and clinical features was constructed, and calibration plots showed that the nomogram resulted in an accurate prognosis for patients. The three-gene signature was associated with T stage, M stage and degree of immune infiltration in patients with cervical cancer.ConclusionsThis research suggests that the developed three-gene signature may be applied as a biomarker to predict the prognosis of and personalized therapy for CESC.

Symptom clusters in COVID-19: A potential clinical prediction tool from the COVID Symptom Study app.

As no one symptom can predict disease severity or the need for dedicated medical support in coronavirus disease 2019 (COVID-19), we asked whether documenting symptom time series over the first few days informs outcome. Unsupervised time series clustering over symptom presentation was performed on data collected from a training dataset of completed cases enlisted early from the COVID Symptom Study Smartphone application, yielding six distinct symptom presentations. Clustering was validated on an independent replication dataset between 1 and 28 May 2020. Using the first 5 days of symptom logging, the ROC-AUC (receiver operating characteristic - area under the curve) of need for respiratory support was 78.8%, substantially outperforming personal characteristics alone (ROC-AUC 69.5%). Such an approach could be used to monitor at-risk patients and predict medical resource requirements days before they are required.

Optimization modelling to establish false measures implemented with ex-situ plant species to control gully erosion in a monsoon-dominated region with novel in-situ measurements

Water dominated gullies formation and associated land degradation are the foremost challenges among the planners for sustainability and optimization of land resources. This type of hazardous phenomenon is utmost vulnerable due to huge loss of surface soil in the sub-tropical developing countries like India. The present study has been carried out in rugged badland topography of Garhbeta-I Community Development (C.D.) Block in eastern India for assessing the gully erosion susceptibility (GES) mapping and optimization of land use planning. The GES mapping is the first and foremost steps towards minimization this adverse affect and attaining sustainable development. In this study we also describe the importance of plantation and alternation of ex-situ tree species with in-situ species for minimizes the erosional activity. To meet our research goal here we used two prediction based machine learning algorithm (MLA) namely random forest (RF) and boosted regression tree (BRT) and one optimization model of Ecogeography based optimization (EBO). The research study also carried out by using a total of 199, in which 139 (70%) and 60 (30%) gully head-cut points were used for training and validation purposes respectively and treated as dependent factors, and twenty gully erosion conditioning factors as independent variables. These models are validated through receiver operating characteristics-area under the curve (ROC-AUC), accuracy (ACC), precision (PRE) and Kappa coefficient index analysis. The validation result showed that EBO model with the highest values of AUC-0.954, ACC-0.85, PRE-0.877 and Kappa-0.646 is the most accurate model for GES followed by BRT and RF. The outcome results should help for the sustainable development of this rugged badland topography.

Metabolic changes in serum steroids for diagnosing and subtyping Cushing’s syndrome

To evaluate the diagnostic value of serum levels of adrenal steroids for diagnosing and subtyping Cushing’s syndrome. Patients diagnosed with endogenous Cushing’s syndrome (34 and 19 patients with adrenal and pituitary Cushing’s syndrome, respectively) and healthy controls (n = 34) were consecutively enrolled at Seoul National University from 2016 to 2020. Morning serum samples were collected before and 3 months after treatment. Serum steroids were profiled using liquid chromatography-mass spectrometry. The diagnostic value of each and the combination of steroids were assessed using the area under the curve of receiver operating characteristic (AUROC) and decision tree analysis. Tetrahydrocortisone and 6β-hydroxycortisol showed the highest AUROC (0.893 and 0.890, respectively) for the diagnosis of endogenous Cushing’s syndrome. The decision tree composed of tetrahydrocortisone and 6β-hydroxycortisol correctly classified 79/87 (90.8 %) subjects. For subtyping into adrenal or pituitary Cushing’s syndrome, dehydroepiandrosterone sulfate (DHEA-S) showed the highest AUROC (0.988), which was similar to that of plasma ACTH (0.994, P = 0.458). The decision tree composed of only DHEA-S correctly classified 51/53 (96.2 %) of the Cushing’s syndrome subtype. DHEA-S showed a significant linear correlation with the plasma ACTH level, but not with the 24 -h urine free cortisol or dexamethasone suppression test results. All steroids, except allo-tetrahydrocortisol and tetrahydrocortisone, decreased significantly at 3 months post-treatment with similar patterns in both adrenal and pituitary Cushing’s syndrome. Serum steroid profiling using a single morning serum sample provides valuable information for diagnosing and subtyping Cushing’s syndrome.

Validation of the IMPROVE‐DD risk assessment model for venous thromboembolism among hospitalized patients with COVID‐19

BackgroundAntithrombotic guidance statements for hospitalized patients with coronavirus disease 2019 (COVID‐19) suggest a universal thromboprophylactic strategy with potential to escalate doses in high‐risk patients. To date, no clear approach exists to discriminate patients at high risk for venous thromboembolism (VTE). ObjectivesThe objective of this study is to externally validate the IMPROVE‐DD risk assessment model (RAM) for VTE in a large cohort of hospitalized patients with COVID‐19 within a multihospital health system. MethodsThis retrospective cohort study evaluated the IMPROVE‐DD RAM on adult inpatients with COVID‐19 hospitalized between March 1, 2020, and April 27, 2020. Diagnosis of VTE was defined by new acute deep venous thrombosis or pulmonary embolism by Radiology Department imaging or point‐of‐care ultrasound. The receiver operating characteristic (ROC) curve was plotted and area under the curve (AUC) calculated. Sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) were calculated using standard methods. ResultsA total of 9407 patients were included, with a VTE prevalence of 2.9%. The VTE rate was 0.4% for IMPROVE‐DD score 0‐1 (low risk), 1.3% for score 2‐3 (moderate risk), and 5.3% for score ≥ 4 (high risk). Approximately 45% of the total population scored high VTE risk, while 21% scored low VTE risk. IMPROVE‐DD discrimination of low versus medium/high risk showed sensitivity of 0.971, specificity of 0.218, PPV of 0.036, and NPV of 0.996. ROC AUC was 0.702. ConclusionsThe IMPROVE‐DD VTE RAM demonstrated very good discrimination to identify hospitalized patients with COVID‐19 as low, moderate, and high VTE risk in this large external validation study with potential to individualize thromboprophylactic strategies.

A prediction model of enteral nutrition complicated with severe diarrhea in ICU patients based on CD55.

Severe diarrhea is a common complication of enteral nutrition in intensive care unit (ICU) patients. CD55 not only plays a vital role in immune but also plays a crucial role in intestinal function. We intended to build a prediction model of enteral nutrition complicated with severe diarrhea in ICU patients based on CD55. This was a prospective, single-center, observational study. We collected 116 patients with enteral nutrition in the ICU. We collected blood samples from patients at the time of admission, tested blood biomarkers [CD55, interleukin-10 (IL-10), diamine oxidase, D-lactic acid and endotoxin], and recorded daily defecation and enteral nutrition. Finally, through multi-factor logistic regression model, a prediction model based on multiple prediction indicators was formed, and new joint predictive factors were calculated. The prediction model of enteral nutrition complicated with severe diarrhea in ICU patients was constructed through data processing analysis. A total of 116 adult patients with enteral nutrition were divided into two groups: 77 patients without severe diarrhea and 39 with severe diarrhea. Compared with patients without severe diarrhea, CD55 on granulocyte membrane surface(gCD55) of patients with severe diarrhea was significantly reduced (P<0.0001), and CD55 on erythrocyte membrane surface(eCD55) of patients with diarrhea was significantly reduced (P<0.0001), serum IL-10 in patients with diarrhea significantly decreased compared with those without diarrhea, P<0.0001.The area under the curve of receiver operating characteristic (AUCROC) was 0.919 in gCD55 predicting severe diarrhea, when cutoff value =99.95%, the optimal sensitivity was 96.2%, and specificity was 100%. The AUROC of eCD55 and IL-10 in predicting severe diarrhea was 0.658 and 0.558 respectively. Based on this, a prediction model was established. The prediction model was: In[P/(1- P)] =-24.68 gCD55 -2.26 eCD55 +0.0041 IL-10 +2,691, the goodness of fit test of the model was P=0.44, and the AUCROC of the model was 0.92. gCD55 and eCD55 had certain predictive value in enteral nutrition complicated with severe diarrhea.

Detection of arterial pressure waveform error using machine learning trained algorithms

In critically ill and high-risk surgical room patients, an invasive arterial catheter is often inserted to continuously measure arterial pressure (AP). The arterial waveform pressure measurement, however, may be compromised by damping or inappropriate reference placement of the pressure transducer. Clinicians, decision support systems, or closed-loop applications that rely on such information would benefit from the ability to detect error from the waveform alone. In the present study we hypothesized that machine-learning trained algorithms could discriminate three types of transducer error from accurate monitoring with receiver operator characteristic (ROC) curve areas greater than 0.9. After obtaining written consent, patient arterial line waveform data was collected in the operating room in real-time during routine surgery requiring arterial pressure monitoring. Three deliberate error conditions were introduced during monitoring: Damping, Transducer High, and Transducer Low. The waveforms were split up into 10s clips that were featurized. The data was also either calibrated against the patient's own baseline or left uncalibrated. The data was then split into training and validation sets, and machine-learning algorithms were run in a Monte-Carlo fashion on the training data with variable sized training sets and hyperparameters. The algorithms with the highest balanced accuracy were pruned, then the highest performing algorithm in the training set for each error state (High, Low, Damped) for both calibrated and uncalibrated data was finally tested against the validation set and the ROC and precision-recall curve area-under the curve (AUC) calculated. 38 patients were enrolled in the study with a mean age of 52 ± 15years. A total of 40h of monitoring time was recorded with approximately 120,000 heart beats featurized. For all error states, ROC AUCs for algorithm performance on classification of the state were greater than 0.9; when using patient-specific calibrated data AUCs were 0.94, 0.95, and 0.99 for the transducer low, transducer high, and damped conditions respectively. Machine-learning trained algorithms were able to discriminate arterial line transducer error states from the waveform alone with a high degree of accuracy.

Mapping method of rainfall-induced landslide hazards by infiltration and slope stability analysis

By utilizing the Green-Ampt infiltration equation and the infinite slope stability model, a method for analyzing shallow slope failures caused by rainfall is developed. With rainfall intensity, soil characteristics, and topography, the modified Green-Ampt infiltration equation is used to estimate the rainfall infiltration capacity and depth of infiltration in a given slope. Assigning the calculated depth of infiltration as the depth of slip surface, the factor of safety of the slope is obtained through the infinite slope stability model. A time-series visualization map of the space-time varying factor of safety is generated when the method is implemented with the aid of Geographic Information System (GIS) software. The model is applied and validated with the landslides that occurred during the October 2019 Typhoon Hagibis in Marumori, Japan. The model results show good agreement with the reported time and depths of failure, and the analysis of the spatial distribution of predicted failures yielded receiver operating characteristic-area under the curve (ROC-AUC) value of 0.90. The applicability of the model can be extended for post-analysis, real-time, or projected assessment of slope stability, depending on the nature of input rainfall data (e.g., historical, real-time, forecast, hypothetical).

A multiple sclerosis disease progression measure based on cumulative disability

Background: Existing severity measurements in multiple sclerosis (MS) are often cross-sectional, making longitudinal comparisons of disease course between individuals difficult. Objective: The objective of this study is to create a severity metric that can reliably summarize a patient’s disease course. Methods: We developed the nARMSS – normalized ARMSS (age-related MS severity score) over follow-up, using the deviation of individual ARMSS scores from the expected value and integrated over the corresponding time period. The nARMSS scales from −5 to +5; a positive value indicates a more severe disease course for a patient when compared to other patients with similar disease timings. Results: Using Swedish MS registry data, the nARMSS was tested using data at 2 and 4 years of follow-up to predict the most severe quartile during the subsequent period up to 10 years total follow-up. The metric used was area under the curve of the receiver operating characteristic (AUC-ROC). This resulted in measurements of 0.929 and 0.941. In an external Canadian validation cohort, the equivalent AUC-ROCs were 0.901 and 0.908. Conclusion: The nARMSS provides a reliable, generalizable and easily measurable metric which makes longitudinal comparison of disease course between individuals feasible.

Deep learning for computer-assisted diagnosis of hereditary diffuse gastric cancer.

BackgroundPatients with hereditary diffuse gastric cancer often undergo prophylactic gastrectomy to minimize cancer risk. Because intramucosal poorly cohesive carcinomas in this setting are typically not grossly visible, many pathologists assess the entire gastrectomy specimen microscopically. With 150 or more slides per case, this is a major time burden for pathologists. This study utilizes deep learning methods to analyze digitized slides and detect regions of carcinoma.MethodsProphylactic gastrectomy specimens from seven patients with germline CDH1 mutations were analyzed (five for training/validation and two for testing, with a total of 133 tumor foci). All hematoxylin and eosin slides containing cancer foci were digitally scanned, and patches of size 256 × 256 pixels were randomly extracted from regions of cancer as well as from regions of normal background tissue, resulting in 15,851 images for training/validation and 970 images for testing. A model with DenseNet-169 architecture was trained for 150 epochs, then evaluated on images from the test set. External validation was conducted on 814 images scanned at an outside institution.ResultsOn individual patches, the trained model achieved a receiver operating characteristic (ROC) area under the curve (AUC) of 0.9986. This enabled it to maintain a sensitivity of 90% with a false-positive rate of less than 0.1%. On the external validation dataset, the model achieved a similar ROC AUC of 0.9984. On whole slide images, the network detected 100% of tumor foci and correctly eliminated an average of 99.9% of the non-cancer slide area from consideration.ConclusionsOverall, our model shows encouraging progress towards computer-assisted diagnosis of hereditary diffuse gastric cancer.

Association between different scale bars in dermoscopic images and diagnostic performance of a market-approved deep learning convolutional neural network for melanoma recognition

BackgroundStudies systematically unravelling possible causes for false diagnoses of deep learning convolutional neural networks (CNNs) are scarce, yet needed before broader application. ObjectivesThe objective of the study was to investigate whether scale bars in dermoscopic images are associated with the diagnostic accuracy of a market-approved CNN. MethodsThis cross-sectional analysis applied a CNN trained with more than 150,000 images (Moleanalyzer-pro®, FotoFinder Systems Inc., Bad Birnbach, Germany) to investigate seven dermoscopic image sets depicting the same 130 melanocytic lesions (107 nevi, 23 melanomas) without or with digitally superimposed scale bars of different manufacturers. Sensitivity, specificity and area under the curve (AUC) of receiver operating characteristics (ROC) for the CNN's binary classification of images with or without superimposed scale bars were assessed. ResultsSix dermoscopic image sets with different scale bars and one control set without scale bars (overall 910 images) were submitted to CNN analysis. In images without scale bars, the CNN attained a sensitivity [95% confidence interval] of 87.0% [67.9%–95.5%] and a specificity of 87.9% [80.3%–92.8%]. ROC AUC was 0.953 [0.914–0.992]. Scale bars were not associated with significant changes in sensitivity (range 87%–95.7%, all p ≥ 1.0). However, four scale bars induced a decrease of the CNN's specificity (range 0%–43.9%, all p < 0.001). Moreover, ROC AUC was significantly reduced by two scale bars (range 0.520–0.848, both p ≤ 0.042). ConclusionsSuperimposed scale bars in dermoscopic images may impair the CNN's diagnostic accuracy, mostly by increasing the rate of the false-positive diagnoses. We recommend avoiding scale bars in images intended for CNN analysis unless specific measures counteracting effects are implemented. Clinical trial numberThis study was registered at the German Clinical Trial Register (DRKS-Study-ID: DRKS00013570; URL: https://www.drks.de/drks_web/).

Alterations of serum levels of plasminogen, TNF-α, and IDO in granulomatosis with polyangiitis patients.

Granulomatosis with polyangiitis (GPA) is a representative of vasculitides associated with anti-neutrophil cytoplasmic autoantibodies. "Classical" antibodies directed against proteinase 3 are involved in the pathogenesis and are part of the GPA diagnosis at the same time. Along with them, however, antibodies against Lysosomal-Associated Membrane Protein-2 (LAMP-2) and antibodies directed against plasminogen have been described in GPA.Objectives and methodology: We performed a cross-sectional study enrolling 34 patients diagnosed with GPA. Our study was aimed at looking for correlations between serum levels of LAMP-2 and plasminogen and the clinical manifestations of the GPA. Furthermore, we examined serum levels of tumor necrosis factor-alpha (TNF-α) and its associated indoleamine-pyrrole 2,3-dioxygenase (IDO), as well as we looked for a correlation between these cytokines and the clinical manifestations of GPA. The results showed that in GPA, serum plasminogen levels were negatively associated with renal involvement (receiver operating characteristic (ROC) area under the curve (AUC) of 0.78) (95% CI 0.53-0.91), p = 0.035, and the extent of proteinuria, Spearman's Rho = -0.4, p = 0.015. Increased levels of TNF-α and IDO correlated with disease activity, Spearman's Rho =0.62, p = 0.001 and Spearman's Rho = 0.4, p = 0.022, respectively, whereas only TNF-α was increased in severe forms of GPA with lung involvement (ROC AUC of 0.8) (95% CI 0.66-0.94), p = 0.005. In this study, we demonstrate the alteration of soluble factors, which play an important role in the pathogenesis of GPA and their relationship with the clinical manifestations of the disease. Our main results confirm the associations of increased secretory TNF-α and some clinical manifestations, and we describe for the first time decreased serum plasminogen levels and their association with renal involvement.

Value of Perioperative Chest X-ray for the Prediction of Sternal Wound Complications after Cardiac Surgery in High-Risk Patients: A "Work in Progress" Analysis.

Background. Sternal wound complications are serious events that occur after cardiac surgery. Few studies have investigated the predictive value of chest X-ray radiological measurements for sternal complications. Methods. Several perioperative radiological measurements at chest X-ray and clinical characteristics were computed in 849 patients deemed at high risk for sternal dehiscence (SD) or More than Grade 1 Surgical Site Infection (MG1-SSI). Multivariable analysis identified independent predictors, whilst receiver operating characteristics (ROC) curve analyses highlighted cut-off values of radiological measurements for the prediction of both complications. Results. SD occurred in 8.8% of the patients, MG1-SSI in 6.8%. Chronic obstructive pulmonary disease (COPD) was the only independent predictor for SD (Odds Ratio, O.R. 12.1; p < 0.001); proximal sternal height (PSH) was the only independent protective factor (O.R. 0.58; p < 0.001), with a cut-off value of 11.7 mm (sensitivity 70.5%, specificity 71.0%; ROC area under the curve (AUC) = 0.768, p < 0.001). Diabetes mellitus (O.R. 3.5; p < 0.001) and COPD (O.R. 21.3; p < 0.001) were independent predictors for MG1-SSI; indexed proximal sternal height (iPSH) was as a protective factor (O.R. 0.26; p < 0.001) with a cut-off of 5.97 mm (sensitivity 70.2%, specificity 69.0%; ROC AUC = 0.739, p < 0.001). No other radiological measurements were independently correlated with SD or MG1-SS (p = N.S.). Conclusion. PSH and iPSH at preoperative chest X-ray may act as indicators of high risk for sternal wound complications, allowing for early preventative measures.

Development of a Predictive Model of Intraoperative Conversion to Thoracotomy for Lung Cancer Patients Undergoing Video-Assisted Thoracoscopic Surgery Lobectomy

Background: The predictors of intraoperative conversion to thoracotomy (ICT) remained inadequately understood. This multicenter trial aimed to develop a predictive model based on logistic regression model and other machine learning algorithms, thus to evaluate the risk of ICT for lung cancer patients undergoing video-assisted thoracoscopic surgery (VATS) lobectomy. Methods: We retrospectively collected data of lung cancer patients undergoing VATS lobectomy from three centers from September 2005 to August 2019. Predictors were selected based on stepwise selection, least absolute shrinkage and selection operator regression. Modelling approach included logistic regression, artificial neural network and random forest. Models was compared with area under the curve (AUC) of receiver operating characteristic (ROC), brier score and Logarithmic-loss in internal validation (10-fold cross validation) and external validation (200 bootstraps). Findings: Finally, 6546 patients were included, and 251 (3.8%) patients experienced ICT. Logistic regression-based model with nine predictors consisting of age, sex, peak expiratory flow, tumor size, clinical TNM stages, VATS ports, resection lobes, with/without pleural adhesion and with/without well-developed interlobar lung fissure, were selected based on the best discrimination and calibration in internal (AUC of ROC: 0.709, 95% CI: 0.560-0.806; Brier score: 0.036, 95%CI: 0.017-0.078; Logarithmic-loss: 0.154, 95%CI: 0.087-0.307) and external validation (AUC of ROC: 0.759, 95% CI: 0.519-0.968; Brier score: 0.037, 95% CI: 0.033-0.041; Logarithmic-loss: 0.152, 95% CI: 0.124-0.179). Interpretation: Logistic-regression based predictive model with nine selected predictors had the capability to evaluate the risk of ICT for lung cancer patients undergoing VATS lobectomy. Larger prospective study was needed to validate model performance. Funding Information: This work was supported by 1.3.5 Project for Disciplines of Excellence, West China Hospital, Sichuan University (ZYGD18021, ZYJC18009), and Post-Doctor Research Project, West China Hospital, Sichuan University (2020HXBH109). Declaration of Interests: The authors declare no potential conflicts of interest. Ethics Approval Statement: The Ethics Committee of West China Hospital of Sichuan University (No. 2020-1034) and Affiliated Hospital of Zunyi Medical University (No. KLL-2020-278) approved this study and waived the requirement of informed consent.

Machine Learning-Based Prediction of COVID-19 Severity and Progression to Critical Illness Using CT Imaging and Clinical Data.

ObjectiveTo develop a machine learning (ML) pipeline based on radiomics to predict Coronavirus Disease 2019 (COVID-19) severity and the future deterioration to critical illness using CT and clinical variables.Materials and MethodsClinical data were collected from 981 patients from a multi-institutional international cohort with real-time polymerase chain reaction-confirmed COVID-19. Radiomics features were extracted from chest CT of the patients. The data of the cohort were randomly divided into training, validation, and test sets using a 7:1:2 ratio. A ML pipeline consisting of a model to predict severity and time-to-event model to predict progression to critical illness were trained on radiomics features and clinical variables. The receiver operating characteristic area under the curve (ROC-AUC), concordance index (C-index), and time-dependent ROC-AUC were calculated to determine model performance, which was compared with consensus CT severity scores obtained by visual interpretation by radiologists.ResultsAmong 981 patients with confirmed COVID-19, 274 patients developed critical illness. Radiomics features and clinical variables resulted in the best performance for the prediction of disease severity with a highest test ROC-AUC of 0.76 compared with 0.70 (0.76 vs. 0.70, p = 0.023) for visual CT severity score and clinical variables. The progression prediction model achieved a test C-index of 0.868 when it was based on the combination of CT radiomics and clinical variables compared with 0.767 when based on CT radiomics features alone (p < 0.001), 0.847 when based on clinical variables alone (p = 0.110), and 0.860 when based on the combination of visual CT severity scores and clinical variables (p = 0.549). Furthermore, the model based on the combination of CT radiomics and clinical variables achieved time-dependent ROC-AUCs of 0.897, 0.933, and 0.927 for the prediction of progression risks at 3, 5 and 7 days, respectively.ConclusionCT radiomics features combined with clinical variables were predictive of COVID-19 severity and progression to critical illness with fairly high accuracy.

Usefulness of the blood urea nitrogen-to-serum albumin ratio as a prognostic indicator of severity in acute ischemic stroke

Objectives: Acute ischemic stroke (AIS) is one of the most important and major causes of mortality worldwide. In AIS patients, the blood urea nitrogen to creatinine ratio (B/C ratio) was investigated as a dehydrated biomarker in previous studies. However, the B/C ratio can be affected by medications and past medical history. We hypothesized addition of serum albumin, which has been shown to have beneficial neuroprotective effects, could compensate for the disadvantages. In the present study, the BUN to serum albumin ratio (B/A ratio) was evaluated association with AIS patient’s prognosis. Methods: This retrospective cohort study of AIS in our hospital was conducted from February 2018 through June 2020. First, demographic and clinical data were collected and compared with the prevalence of mortality and ICU admission. Then, the B/C ratio and the B/A ratio were calculated. Differences in the performance between the B/C ratio and the B/A ratio for outcome prediction were evaluated based on the area under the curve of the receiver operating characteristic (AUROC). Results: Among the 1,164 patients enrolled in this study, 77 (6.6%) died during hospitalization and 467 (40.1%) were admitted to ICU. Multivariate logistic regression had shown that the B/A ratio was a significant predictor of mortality and admission to ICU. In addition, the B/A ratio was significantly higher in ICU-admitted patients and non-survivors. The AUROC of the B/A ratio was 0.687 and the B/C ratio was 0.533 for predicting mortality; the B/A ratio was statistically superior to the B/C ratio. For predicting ICU admission, the AUROC values of the B/A ratio and the B/C ratio were 0.567 and 0.556, respectively, and a statistically significant difference was not observed. Conclusion: The B/A ratio is a simple and useful tool for predicting the outcomes of ischemic stroke patients.

Diagnostic accuracy of the cancer ratio for the prediction of malignant pleural effusion: evidence from a validation study and meta-analysis

Objective This study aimed to assess the diagnostic accuracy of serum LDH to pleural ADA ratio (cancer ratio, CR)for malignant pleural effusion (MPE) through an original study and meta-analysis. Methods We retrospectively collected data from 145 patients with MPE and 117 cases of benign pleural effusions (BPE). The diagnostic performance of CR and a typical biomarker of MPE, carcinoembryonic antigen (CEA), were analysed using the receiver operating characteristic (ROC) curves and the area under the curve (AUC) as a measure of accuracy. The overall diagnostic accuracy of CR was summarised by a standard diagnostic meta-analysis. Results Significantly higher CR and pleural CEA values were observed in the MPE patients than in the BPE patients. At a cut-off value of 14.97, CR showed high sensitivity (0.91), low specificity (0.67), and high AUC (0.85). The combination of CEA and CR increased the AUC to 0.98. The meta-analysis included seven studies involving 2,078 patients. The pooled values for sensitivity, specificity, positive/negative likelihood ratio, and diagnostic odds ratio of CR were 0.96, 0.88, 7.70, 0.05, and 169, respectively. The AUC of the summary ROC of CR was 0.98. Conclusion CR has a high diagnostic accuracy for predicting MPE, especially when used in combination with pleural CEA.