Validation Subset Research Articles

Identification of Alzheimer's disease using a convolutional neural network model based on T1-weighted magnetic resonance imaging

The classification of Alzheimer’s disease (AD) using deep learning methods has shown promising results, but successful application in clinical settings requires a combination of high accuracy, short processing time, and generalizability to various populations. In this study, we developed a convolutional neural network (CNN)-based AD classification algorithm using magnetic resonance imaging (MRI) scans from AD patients and age/gender-matched cognitively normal controls from two populations that differ in ethnicity and education level. These populations come from the Seoul National University Bundang Hospital (SNUBH) and Alzheimer’s Disease Neuroimaging Initiative (ADNI). For each population, we trained CNNs on five subsets using coronal slices of T1-weighted images that cover the medial temporal lobe. We evaluated the models on validation subsets from both the same population (within-dataset validation) and other population (between-dataset validation). Our models achieved average areas under the curves of 0.91–0.94 for within-dataset validation and 0.88–0.89 for between-dataset validation. The mean processing time per person was 23–24 s. The within-dataset and between-dataset performances were comparable between the ADNI-derived and SNUBH-derived models. These results demonstrate the generalizability of our models to different patients with different ethnicities and education levels, as well as their potential for deployment as fast and accurate diagnostic support tools for AD.

Two-stage variational mode decomposition and support vector regression for streamflow forecasting

Abstract. Streamflow forecasting is a crucial component in the management and control of water resources. Decomposition-based approaches have particularly demonstrated improved forecasting performance. However, direct decomposition of entire streamflow data with calibration and validation subsets is not practical for signal component prediction. This impracticality is due to the fact that the calibration process uses some validation information that is not available in practical streamflow forecasting. Unfortunately, independent decomposition of calibration and validation sets leads to undesirable boundary effects and less accurate forecasting. To alleviate such boundary effects and improve the forecasting performance in basins lacking meteorological observations, we propose a two-stage decomposition prediction (TSDP) framework. We realize this framework using variational mode decomposition (VMD) and support vector regression (SVR) and refer to this realization as VMD-SVR. We demonstrate experimentally the effectiveness, efficiency and accuracy of the TSDP framework and its VMD-SVR realization in terms of the boundary effect reduction, computational cost, and overfitting, in addition to decomposition and forecasting outcomes for different lead times. Specifically, four comparative experiments were conducted based on the ensemble empirical mode decomposition (EEMD), singular spectrum analysis (SSA), discrete wavelet transform (DWT), boundary-corrected maximal overlap discrete wavelet transform (BCMODWT), autoregressive integrated moving average (ARIMA), SVR, backpropagation neural network (BPNN) and long short-term memory (LSTM). The TSDP framework was also compared with the wavelet data-driven forecasting framework (WDDFF). Results of experiments on monthly runoff data collected from three stations at the Wei River show the superiority of the VMD-SVR model compared to benchmark models.

Abstract 364: Continuous Electrocardiogram Classification During Resuscitation Using Transfer Learning

Introduction: Cardiac arrest resuscitation requires CPR interruption for ECG rhythm analysis, but pausing CPR is adversely associated with survival. Ideally, automated rhythm analysis would occur agnostic of CPR state throughout resuscitation and discriminate non-shockable from shockable rhythms. Transfer learning of pre-trained deep convolutional neural networks (CNNs) may enable accurate ECG analysis when applied to time-frequency representations of the ECG. We designed and evaluated a transfer learning algorithm to identify ventricular fibrillation (VF), asystole (AS), and organized (OR) rhythms agnostic of CPR. Methods: In this observational study of out-of-hospital cardiac arrest, rhythms were manually diagnosed in continuous defibrillator ECG recordings. Non-overlapping adjacent 2-s ECG segments were extracted from the first 30 min of each case regardless of CPR during VF, AS, and OR. Each segment was represented by an intrafrequency-normalized Morlet wavelet transform from 4-40 Hz. Using a 2/3 subset of patients for training, a series of two ResNet-101 CNNs were retrained to perform a shock decision (VF vs. non-shockable) followed by a specific non-shockable prediction (AS, OR, or Indeterminate). Performance was evaluated in a 1/3 validation subset of patients using a range of probability decision thresholds to predict the class of each segment. Results: In total, 275100 segments were collected from 461 patients. Of 90962 segments from 152 validation patients, using a 0.7 probability threshold for class prediction, 21% (18930/90962) were indeterminate, shock vs. no-shock sensitivity and specificity were 90% (19702/21930) and 97% (48421/50102), and specificities among non-shockable rhythms for AS vs. OR were 84% (5032/5998) and 86% (37760/44104), respectively (Fig 1). Conclusion: Transfer learning may enable shock/no-shock and rhythm-specific ECG classification continuously throughout resuscitation regardless of CPR.

An extended flame index partitioning for partially premixed combustion

Compared to partially premixed combustion (or combustion of non-homogeneous reactants in general), fully premixed and diffusion flames represent only two asymptotic limits of combustion modes. However, the deep knowledge accumulated over the years on these two elementary and archetypal flame prototypes is such that they remain the cornerstone and reference building blocks of most combustion modelling proposals. Therefore, from a general point of view, being able to distinguish between premixed and non-premixed modes of combustion thanks to a flame index appears as a quite appealing but challenging task that still concentrates many research efforts. Indeed, the availability of such an index is not only appealing to proceed with the analysis of either experimental or computational data issued from DNS (or highly resolved LES) databases. It is also an essential ingredient to elaborate advanced flamelet-based multiregime combustion models on the basis of single regime tabulated flamelet databases. In the present study, a new definition of the premixedness index is proposed for partially premixed combustion. It is based on a weighted form of the cross-scalar dissipation rate of the mixture fraction and progress variable , i.e. quantities that have been previously identified as relevant parameters to describe partially premixed combustion regimes. The relevance of the corresponding index is assessed through a detailed computational procedure that includes three successive validation subsets: counterflow flames (including premixed, rich partially-premixed, and diffusion flames), (ii) stabilised triple flames for three distinct values of the inlet mixture fraction gradient, and finally (iii) unsteady flame kernel developments in non-homogeneous mixtures of fresh reactants, which are characterised by various initial levels of the segregation rate between the fuel and oxidiser. The proposed premixedness index and its counterpart are used as the weighting coefficients between tabulated premixed flamelets (TPF) and tabulated diffusion flamelets (TDF) data, which have been parameterised as functions of and . It is noteworthy that, in contrast to some previous proposals of the literature, the present flame index does not require the consideration of any other quantities in addition to those already used to parameterise the flamelets databases, i.e. and . The validation procedure makes use of steady and unsteady processes with a priori and a posteriori analyses. In both cases, the comparisons between the results obtained with the proposed flame partitioning and detailed chemistry (DC) computations lead to a satisfactory level of agreement and, from a general viewpoint, the level of agreement is better than the one obtained with either premixed or diffusion flamelet-based models.

Artificial Neural Networks as a Way to Predict Future Kidney Cancer Incidence in the United States.

The incidence of kidney cancer is increasing; it could be counteracted with new ways to predict and detect it. We aimed to implement an artificial neural network in order to predict new cases of renal-cell carcinoma (RCC) in the population using population rate, obesity, smoking incidence, uncontrolled hypertension, and life expectancy data in the United States. Statistics were collected on US population numbers, life expectancy, obesity, smoking, and hypertension. We used MATLAB R2018 (MathWorks) software to implement an artificial neural network. Data were repeatedly and randomly divided into training (70%) and validation (30%) subsets. The number of new RCC cases will grow from 44,400 (2020) to 55,400 (2050), an increase of+24.7%. Our data show that preventing hypertension would have the greatest impact on reduction of the incidence, estimated at-775 and-575 cases per year in 2020 and in 2030, respectively. The prevention of obesity and smoking would have a more limited impact, estimated at-64 and-180 cases per year in 2020 and in 2030, respectively, for obesity, and-173 and-21 cases per year in 2020 and in 2030, respectively, for smoking. Our predictions underline the need for accurate studies on RCC-related risk factors to reduce the incidence.

Post-Treatment Clone Size Predicts Survival Independently of IPSS-R and Response after Azacitidine Therapy for MDS.

Background DNA hypomethylating agents (HMAs), including azacytidine (AZA) have been established as key drugs for higher-risk myelodysplastic syndromes (MDS). We and others have explored the role of mutation profile before AZA administration on predicting outcomes. Actually, we have previously identified mutated-TP53 as a marker associated with higher rate of achieving complete remission (CR). In addition, mutations in TP53 and DDX41 predicted reduced and prolonged survival after treatment, respectively. However, the clinical significance of evaluating clone size changes early after treatment has not been determined. In this study, we explored the role of post-treatment clone size in predicting outcomes of AZA treatment for MDS and related diseases. Methods We enrolled 290 AZA-treated cases, including 88 from a Japanese prospective study (JALSG MDS-212 trial), 149 from Karolinska Institute, and 53 from a retrospectively collected Japanese cases. The diagnoses were MDS (n=242), MDS/MPN (n=25), and AML-MRC (n=23). For all patients, tumor samples were collected both before and after AZA administration and were analyzed for mutations in 66 genes implicated in myeloid neoplasms using targeted-capture sequencing. The median cycles of AZA treatment before sampling was 4 (range 1-7). Clone size was calculated from variant allele frequency adjusted for ploidy or allelic imbalances.Survival was calculated with a Cox regression model. Results In post-treatment samples, we identified 870 mutations in 51 genes in 255 (88%) patients with a median of 3 mutations per sample, while 943 mutations were seen in 279 (96%) patients in the pre-treatment samples. Most frequently detected mutations in post-treatment samples were seen in TET2, TP53, RUNX1, and ASXL1. Germline DDX41 mutations were excluded from clone size evaluation. Median clone sizes were 0.63 and 0.54 for pre-treatment and post-treatment samples (P=.011), respectively. The largest clone sizes (max(VAF)) in post-treatment samples had a strong negative correlation with hematological response according to IWG criteria (P < .0001). We next explored whether max(VAF) in post-treatment samples provides a more precise estimation of long-term survival than IPSS-R. Max(VAF) further stratified each IPSS-R risk group in subgroups with discrete OS (P < .0001 for IPSS-R very high and P = .0004 for high risk group). Incorporating pre-treatment mutation data (mutations in TP53 and DDX41) and max(VAF) values in addition to IPSS-R scores and clinical response, we constructed a multivariate model and found that all these factors had an independent and significant impact on OS (Figure 1A). Next, we examined whether max(VAF) combined with IPSS-R and clinical response can improve the model. For this purpose, we randomly split the cohort into 75% training and 25% validation subsets and for each split, we constructed different models using the training set, performance of which was evaluated by calculating the concordance index (c-index) using the validation set. The mean c-index in 10,000 simulation sets increased by 0.025 by adding response data to IPSS-R score (I versus IR in Fig 1B). Further improvements were obtained by adding gene mutation and max(VAF), in which the c-index increased by 0.034 (IR versus IGR in Fig 1B) and 0.010 (IGR versus IGRP in Fig 1B), respectively. For the 53 patients who received allogeneic stem cell transplantation, the median post-transplant OS was 82.6 months (range, 36.3 to not reached). Notably, max(VAF) significantly stratified OS after allo-SCT (HR, 3.3; 95%CI, 1.3 to 8.3; P = .014). Conclusions Our study revealed that post-treatment clone size significantly correlated with clinical response and the evaluation of post-treatment clone size allows for more precise prognostication after AZA treatment compared with IPSS-R and clinical response alone. Table Disclosures Naoe: NIPPON SHINYAKU CO.,LTD.: Speakers Bureau; Sysmex co.: Speakers Bureau; Eisai Co., Ltd.: Speakers Bureau; Astellas Pharma Inc.: Speakers Bureau; Bristol-Myers Squibb Company: Speakers Bureau. Miyazaki:Celgene: Honoraria; Sumitomo Dainippon Pharma Co., Ltd.: Honoraria; Kyowa Kirin Co., Ltd.: Honoraria; Novartis Pharma KK: Honoraria; NIPPON SHINYAKU CO.,LTD.: Honoraria; Otsuka Pharmaceutical: Honoraria; Astellas Pharma Inc.: Honoraria; Chugai Pharmaceutical Co., Ltd.: Honoraria. Papaemmanuil:Kyowa Hakko Kirin: Consultancy, Honoraria; Prime Oncology: Consultancy, Honoraria; Novartis: Consultancy, Honoraria; Illumina: Consultancy, Honoraria; Celgene: Consultancy, Honoraria, Research Funding; MSKCC: Patents & Royalties; Isabl: Current equity holder in private company, Membership on an entity's Board of Directors or advisory committees. Ogawa:Eisai Co., Ltd.: Research Funding; Chordia Therapeutics, Inc.: Membership on an entity's Board of Directors or advisory committees, Research Funding; KAN Research Institute, Inc.: Membership on an entity's Board of Directors or advisory committees, Research Funding; Otsuka Pharmaceutical Co., Ltd.: Research Funding; Asahi Genomics Co., Ltd.: Current equity holder in private company; Sumitomo Dainippon Pharma Co., Ltd.: Research Funding.

Deep learning analysis provides accurate COVID-19 diagnosis on chest computed tomography

IntroductionComputed Tomography is an essential diagnostic tool in the management of COVID-19. Considering the large amount of examinations in high case-load scenarios, an automated tool could facilitate and save critical time in the diagnosis and risk stratification of the disease. MethodsA novel deep learning derived machine learning (ML) classifier was developed using a simplified programming approach and an open source dataset consisting of 6868 chest CT images from 418 patients which was split into training and validation subsets. The diagnostic performance was then evaluated and compared to experienced radiologists on an independent testing dataset. Diagnostic performance metrics were calculated using Receiver Operating Characteristics (ROC) analysis. Operating points with high positive (>10) and low negative (<0.01) likelihood ratios to stratify the risk of COVID-19 being present were identified and validated. ResultsThe model achieved an overall accuracy of 0.956 (AUC) on an independent testing dataset of 90 patients. Both rule-in and rule out thresholds were identified and tested. At the rule-in operating point, sensitivity and specificity were 84.4 % and 93.3 % and did not differ from both radiologists (p > 0.05). At the rule-out threshold, sensitivity (100 %) and specificity (60 %) differed significantly from the radiologists (p < 0.05). Likelihood ratios and a Fagan nomogram provide prevalence independent test performance estimates. ConclusionAccurate diagnosis of COVID-19 using a basic deep learning approach is feasible using open-source CT image data. In addition, the machine learning classifier provided validated rule-in and rule-out criteria could be used to stratify the risk of COVID-19 being present.

Assessing Genitourinary Cancer Clinical Trial Accrual Sufficiency Using Archived Trial Data.

Clinical trials often fail to reach their anticipated end points, most frequently because of poor accrual. Prior studies have analyzed trial termination, but it has not been easy to assess accrual estimates using international databases such as ClinicalTrials.gov because of limitations in accessing accrual information. Specifically, it is not easy to extract both anticipated and actual accrual of clinical trials. We designed a new algorithmic approach to extracting trial accrual data from ClinicalTrials.gov and used it to estimate the sufficiency of patient accrual onto genitourinary (GU) cancer trials. We queried ClinicalTrials.gov for completed/terminated phase II and III clinical trials for prostate, bladder, kidney, testicular, and ureteral cancers registered after 2007. We extracted trial characteristics from available XML files. We then used a Python algorithm to access prior trial registrations on the ClinicalTrials.gov archive site and extract both anticipated and actual accrual numbers. We then compared the actual accrual of each trial to its anticipated accrual and defined sufficient accrual as 85% of anticipated accrual. The algorithm was 100% accurate compared with hand extraction in a small validation subset. A total of 925 trials were included, of which 840 (91%) had both anticipated and actual accrual. Only 418 (50%) trials had sufficient accrual (≥ 85% of anticipated). Considering only trials marked as successfully completed, 395/597 (66%) reached sufficient accrual. GU cancer trials often do not meet their anticipated accrual goals. New approaches to trial conduct are direly needed. Our reproducible and scalable approach to extracting accrual information can be applied to analysis of ClinicalTrials.gov in future analyses in the hope of improving the efficiency of the clinical trials enterprise.

Development of standardized regression-based formulas to assess meaningful cognitive change in early Parkinson's disease.

Longitudinal assessment of cognitive and emotional functioning in patients with Parkinson's disease (PD) is helpful in tracking progression of the disease, developing treatment plans, evaluating outcomes, and educating patients and families. Determining whether change over time is meaningful in neurodegenerative conditions, such as PD, can be difficult as repeat assessment of neuropsychological functioning is impacted by factors outside of cognitive change. Regression-based prediction formulas are one method by which clinicians and researchers can determine whether an observed change is meaningful. The purpose of the current study was to develop and validate regression-based prediction models of cognitive and emotional test scores for participants with early-stage idiopathic PD and healthy controls (HC) enrolled in the Parkinson's Progression Markers Initiative (PPMI). Participants with de novo PD and HC were identified retrospectively from the PPMI archival database. Data from baseline testing and 12-month follow-up were utilized in this study. In total, 688 total participants were included in the present study (NPD=508; NHC=185). Subjects from both groups were randomly divided into development (70%) and validation (30%) subsets. Early-stage idiopathic PD patients and healthy controls were similar at baseline. Regression-based models were developed for all cognitive and self-report mood measures within both populations. Within the validation subset, the predicted and observed cognitive test scores did not significantly differ, except for semantic fluency. The prediction models can serve as useful tools for researchers and clinicians to study clinically meaningful cognitive and mood change over time in PD.

A Pre-Treatment PET And CT-Derived Model for Progression-Free Survival in Early Stage Non-Small Cell Lung Cancer

The maximum value of the standardized uptake value (SUV) and tumor diameter have been associated with progression after stereotactic body radiotherapy (SBRT) for early stage non-small cell lung cancer (NSCLC). In this work, we explored whether other tumor parameters based on pre-SBRT PETs and CTs predict progression-free survival (PFS) in this patient population. 412 patients treated between 2006 and 2017 were included. Patients had to have PETs and CTs available within three months prior to SBRT start. The median prescription dose was 50Gy in 5 fractions. The planning-CT gross tumor volumes (GTVs) were propagated onto the pre-treatment PETs and CTs using b-spline deformable image registration. PET intensity features (90th percentile, entropy, maximum, mean, peak, robust mean absolute deviation, SD, valley) and CT shape features (compactness, diameter, elliptic axes, elongation, flatness, number of lobules/peaks/spicules, sphericity, surface area, surface to volume ratio, volume) were extracted. Data were split into training and hold-out validation subsets (n = 283, 123; 70%, 30%). In the training subset, the imaging features and six patient characteristics (age, gender, histology, performance status, prior surgery, tumor location) were tested for association with PFS using Cox Proportional Hazards regression with re-sampling (bootstrapping with 1000 samples). Significance was denoted at p≤0.0019 (corrected for 26 tests). A bootstrapped forward-stepwise multivariate analysis was undertaken including only non-strongly correlated predictors (Spearman’s rank, |Rs|<0.70). The most frequently selected model was explored in the validation subset in which model performance was assessed using the c-index and the prediction-stratified high and low risk tertiles (HR, LR) of the observed PFS were compared. Nineteen of the 20 identified candidate predictors were either PET or CT features (p-value range: 3E-9, 1.2E-3). The intra-imaging modality correlation between features was strong (median |Rs|: PET: 0.93; CT: 0.76) and only four features were passed on to multivariate analysis: PET entropy, CT number of peaks, CT major axis, and gender. The most frequently selected model included PET entropy and CT number of peaks; the c-index in the validation subset was 0.77 and the prediction-stratified survival indicated a clear separation between the observed HR and LR: e.g. a PFS of 60% was observed at 12 months in HR vs. 22 months in LR. This PET and CT-based model identified the SUV distribution randomness (entropy) and spiculated tumor pattern on CTs as the most important features in predicting PFS in early stage NSCLC. The associated performance on the hold-out validation subset was good and its use has the potential to further improve the prediction of response to SBRT for this patient population. This model will be used to identify high-risk patients based on the predicted PFS in an upcoming phase II study on adjuvant immunotherapy.

MRI Radiomic Habitats Predict for Radiation Response in Soft Tissue Sarcoma

Favorable tumor response (pathologic necrosis [PN] >90-95%) has been associated with improved outcome in soft tissue sarcoma (STS). STS is a diverse disease with variable response to radiation (RT), therefore tools such as radiomics may improve our prediction of tumor radiosensitivity. Radiomic habitats are analyses of overlaid images, providing insight on the underlying innate biology. We hypothesize that utilizing pretreatment MRI habitat analysis can predict for favorable tumor response in STS. We identified a total of 432 non-metastatic sarcoma patients that were treated with neoadjuvant radiation (50Gy/25 fractions) followed by surgery, from 2001 to 2019. Patients with a pre-RT MRI were included, gross volume was manually segmented on T1-post contrast and T2 STIR sequences. The sequences were dichotomized by signal intensity and each region designated into 1 of 4 habitats (e.g. T1 high/low, T2 high/low), and 154 habitat features were extracted. The cohort was split 2:1 into training and validation subsets balanced by PN, clinical tumor T (cT) and N stage (cN), receiving induction chemotherapy, and age. Feature reduction and machine learning was utilized to identify habitat features predictive of PN rates >90% or >95% at the time of surgery. Overall, patients eligible for analysis (n = 97) had a median age of 63 and tumor size of 11 cm. They most commonly consisted of undifferentiated pleomorphic sarcoma (40%), extremity (80%), and no prior chemotherapy (87%). An 8-feature model, 4 radiomic and 4 clinical (recurrent disease, age, cT stage, cN stage), predicted for PN>90% (training: 72%, ROC 0.746; validation: 65%, ROC: 0.713). Evaluating patients treated with neoadjuvant RT alone (n = 84), a second model with 5 radiomic and 3 clinical features (prior RT, cT and cN stage), had the highest prediction for PN>95% (training: 78%, ROC 0.721; validation: 72%, ROC 0.600). In both the overall cohort and RT only subset, the diversity in the number of habitat volumes within the tumor predicted for tumor response (OR 7.71 and 3.31). Models incorporating pretreatment MRI-habitats and clinical features can be used to predict radiation response in STS. Radiomic features suggestive of habitat diversity within a tumor appear to be associated with radiation response, and highlight the importance of intratumoral heterogeneity when evaluating radiosensitivity. Prospective studies utilizing this model to predict radiation response may identify patients that could benefit from treatment escalation.

Regression calibration to correct correlated errors in outcome and exposure.

Measurement error arises through a variety of mechanisms. A rich literature exists on the bias introduced by covariate measurement error and on methods of analysis to address this bias. By comparison, less attention has been given to errors in outcome assessment and nonclassical covariate measurement error. We consider an extension of the regression calibration method to settings with errors in a continuous outcome, where the errors may be correlated with prognostic covariates or with covariate measurement error. This method adjusts for the measurement error in the data and can be applied with either a validation subset, on which the true data are also observed (eg, a study audit), or a reliability subset, where a second observation of error prone measurements are available. For each case, we provide conditions under which the proposed method is identifiable and leads to consistent estimates of the regression parameter. When the second measurement on the reliability subset has no error or classical unbiased measurement error, the proposed method is consistent even when the primary outcome and exposures of interest are subject to both systematic and random error. We examine the performance of the method with simulations for a variety of measurement error scenarios and sizes of the reliability subset. We illustrate the method's application using data from the Women's Health Initiative Dietary Modification Trial.

Development and validation of coding algorithms to identify patients with incident lung cancer in United States healthcare claims data.

Our aim was to develop and validate a practical US healthcare claims algorithm for identifying incident lung cancer that improves on positive predictive value (PPV) and sensitivity observed in past studies. Patients newly diagnosed with lung cancer in Surveillance, Epidemiology, and End Results (SEER) (gold standard) were linked with Medicare claims. A 5% Medicare "other cancer" sample and noncancer sample served as controls. A split-sample validation approach was used. Rules-based, regression, and machine learning models for developing algorithms were explored. Algorithms were developed in the model building subset. Rules-based algorithms and those with the highest F scores were evaluated in the validation subset. F scores were compared for 1000 bootstrap samples. Misclassification was evaluated by calculating the odds of selection by the algorithm among true positives and true negatives. A practical single-score algorithm derived from a logistic regression model had sensitivity = 78.22% and PPV = 78.50% (F score: 78.36). The algorithm was most likely to misclassify older patients (ages ≥80 years) or with missing data in the SEER registry, shorter follow-up time in Medicare (<3 months), insurance through Veterans Affairs, >1 cancer in SEER, or certain Charlson comorbidities (dementia, chronic pulmonary disease, liver disease, or myocardial infarction). In this dataset, a practical point-based algorithm for identifying incident lung cancer demonstrated significant and substantial improvement (7.9% and 23.9% absolute improvement in sensitivity and PPV, respectively) compared with a current standard.

The optimal cut-off point of vitamin D for pregnancy outcomes using a generalized additive model

Several studies consider vitamin D deficiency as a modifiable risk factor for adverse pregnancy outcomes. However, there is no specific cut-off point for the serum level of this prohormone to identify high-risk pregnant women. This study aimed to determine the thresholds for the circulating levels of 25-hydroxyvitamin D (25(OH)D), associated with adverse pregnancy outcomes, including preterm labor, preeclampsia (PE), and gestational diabetes mellitus (GDM), using a generalized additive model. This is a descriptive and analytical cross-sectional study carried out on the data collected from the Khuzestan Vitamin D Deficiency Screening Program in Pregnancy. Of 1800 pregnant women, referred to the health centers of Masjed-Soleyman and Shushtar (Khuzestan Province, Iran), we used the data of 1763 pregnant women, whose serum vitamin D status during the third trimester of pregnancy was available. The datasets were randomly divided into training (70%) and validation (30%) subsets. The cut-off levels of 25(OH)D were identified for the low, moderate, and high risk of adverse pregnancy outcomes, according to generalized additive models (GAM) with smooth functions in the training data set. Then Generalized Linear Model (GLM), with logit link function was applied in the validation dataset to explore the relationships between the optimal vitamin D classification and adverse pregnancy outcomes after adjusting for the potential confounders. The optimal cut-off levels of 25(OH)D for the high, moderate, and low risk of GDM were ≤16, 16-26, and >26ng/mL, respectively. Also, the optimal cut-off points of 25(OH)D for the high, moderate, and low risk of preterm delivery were ≤15, 15-21, and >21ng/mL, respectively. Finally, the corresponding values for the high, moderate, and low risk of PE were ≤15, 15-23, and >23ng/mL, respectively. The models were well-calibrated, based on the Hosmer-Lemeshow test. Results of the adjusted generalized linear model showed a significant increasing trend in the risk of pregnancy outcomes by decreasing 25(OH)D levels. In the preconception period, a 25(OH)D cut-off level of >15ng/mL is recommended for the prevention of adverse pregnancy outcomes.

Mitochondria-derived methylmalonic acid, a surrogate biomarker of mitochondrial dysfunction and oxidative stress, predicts all-cause and cardiovascular mortality in the general population

BackgroundInherited methylmalonic acidemia is characterized by mitochondrial dysfunction, oxidative stress, and damage of mitochondria-rich organs in children. It is unclear whether methylmalonic acid (MMA) is related to poor prognosis in adults. The study aims to investigate the associations of MMA with all-cause and cause-specific mortality in the general population. MethodsOverall, 23,437 adults from the US National Health and Nutrition Examination Survey (NHANES) were enrolled. NHANES 1999–2004 and 2011–2014 were separately used as primary and validation subsets (median follow-up 13.5 and 2.8 years, respectively). Circulating MMA was measured with gas chromatography/mass spectrophotometry. Hazard ratios (HR) were estimated using weighted Cox regression models. ResultsDuring 163,632 person-years of follow-up in NHANES 1999–2004, 3019 deaths occurred. Compared with participants with MMA <120 nmol/L, those with MMA≥250 nmol/L had increased all-cause and cardiovascular mortality in the multivariable-adjusted model [HR(95%CI), 1.62 (1.43–1.84) and 1.66 (1.22–2.27), respectively]. The association was especially significant among participants with normal cobalamin. MMA remained an independent predictor of all-cause mortality occurring whether within 5-year, 5–10 years, or beyond 10-year of follow-up (each p for trend≤0.007). That association was repeatable in NHANES 2011–2014. Moreover, baseline MMA improved reclassification for 10-year mortality in patients with cardiovascular disease (net reclassification index 0.239, integrated discrimination improvement 0.022), overmatched established cardiovascular biomarkers C-reactive protein or homocysteine. ConclusionsCirculating level of mitochondrial-derived MMA is strongly associated with elevated all-cause and cardiovascular mortality. Our results support MMA as a surrogate biomarker of mitochondrial dysfunction to predict poor prognosis in adults. The biological mechanisms under cardiovascular disease warrant further investigation.

Monitoring chest compression rate in automated external defibrillators using the autocorrelation of the transthoracic impedance.

AimHigh-quality chest compressions is challenging for bystanders and first responders to out-of-hospital cardiac arrest (OHCA). Long compression pauses and compression rates higher than recommended are common and detrimental to survival. Our aim was to design a simple and low computational cost algorithm for feedback on compression rate using the transthoracic impedance (TI) acquired by automated external defibrillators (AEDs).MethodsECG and TI signals from AED recordings of 242 OHCA patients treated by basic life support (BLS) ambulances were retrospectively analyzed. Beginning and end of chest compression series and each individual compression were annotated. The algorithm computed a biased estimate of the autocorrelation of the TI signal in consecutive non-overlapping 2-s analysis windows to detect the presence of chest compressions and estimate compression rate.ResultsA total of 237 episodes were included in the study, with a median (IQR) duration of 10 (6–16) min. The algorithm performed with a global sensitivity in the detection of chest compressions of 98.7%, positive predictive value of 98.7%, specificity of 97.1%, and negative predictive value of 97.1% (validation subset including 207 episodes). The unsigned error in the estimation of compression rate was 1.7 (1.3–2.9) compressions per minute.ConclusionOur algorithm is accurate and robust for real-time guidance on chest compression rate using AEDs. The algorithm is simple and easy to implement with minimal software modifications. Deployment of AEDs with this capability could potentially contribute to enhancing the quality of chest compressions in the first minutes from collapse.

Improving uncertainty analysis in well log classification by machine learning with a scaling algorithm

Uncertainty is an important indicator that can provide the confidence in the predictions. However, most machine learning methods in classification problems are incapable of predicting an accurate probability that matches the observed distribution of each class. To overcome this problem and improve the uncertainty analysis in lithofacies classification, a novel calibration process is performed on the validation subset using scaling algorithms. The validation subset should be separated from the training data in order to reduce overfitting and avoid unwanted biases. Random forest is selected as the machine learning classifier owing to its relative ease of use, and can handle high-dimensional input features quite well. The proposed approach is applied to two real datasets to demonstrate the improvement in the uncertainty quantification. Reliability diagram shows that the calibrated probability is closer to the diagonal truth, compared with the uncalibrated one. In addition to the visual check, a multiclass Brier score of calibrated model is lower than that by uncalibrated system, which can be referred to as a loss function. For the classification results by calibrated and uncalibrated classifiers on the testing subset, an increase of the Matthews correlation coefficient also means that the calibrated model has a better performance.

Can Existing Improvements of Water, Sanitation, and Hygiene (WASH) in Urban Slums Reduce the Burden of Typhoid Fever in These Settings?

Sustained investments in water, sanitation, and hygiene (WASH) have lagged in resource-poor settings; incremental WASH improvements may, nonetheless, prevent diseases such as typhoid in disease-endemic populations. Using prospective data from a large cohort in urban Kolkata, India, we evaluated whether baseline WASH variables predicted typhoid risk in a training subpopulation (n = 28 470). We applied a machine learning algorithm to the training subset to create a composite, dichotomous (good, not good) WASH variable based on 4 variables, and evaluated sensitivity and specificity of this variable in a validation subset (n = 28 470). We evaluated in Cox regression models whether residents of "good" WASH households experienced a lower typhoid risk after controlling for potential confounders. We constructed virtual clusters (radius 50 m) surrounding each household to evaluate whether a prevalence of good WASH practices modified the typhoid risk in central household members. Good WASH practices were associated with protection in analyses of all households (hazard ratio [HR] = 0.57; 95% confidence interval [CI], .37-.90; P = .015). This protection was evident in persons ≥5 years old at baseline (HR = 0.47; 95% CI, .34-.93; P = .005) and was suggestive, though not statistically significant, in younger age groups (HR = 0.61; 95% CI, .27-1.38; P = .235). The level of surrounding household good WASH coverage was also associated with protection (HR = 0.988; 95% CI, .979-.996; P = .004, for each percent coverage increase). However, collinearity between household WASH and WASH coverage prevented an assessment of their independent predictive contributions. In this typhoid-endemic setting, natural variation in household WASH was associated with typhoid risk. If replicated elsewhere, these findings suggest that WASH improvements may enhance typhoid control, short of major infrastructural investments.

Technology choices in public transport planning: A classification framework

Choice of public transport technologies in cities is not straightforward: while the academy focuses on optimization models to determine which modes should a specific city have, policy makers rely on simple recommendations which are based on city population and income. We estimated six types of classification models that could allow for more precise recommendations yet are simple enough to be applied by the authorities. We considered typical variables as population and Gross Domestic Product of cities but also geographic and morphologic characteristics in a database of 400 cities from North and South America. Ordered Probit and Multinomial Logit models were the most accurate, with a success rate over 80% in the validation subset. Among the explanatory variables, city population and GDP per capita were as expected the most significant, but fare integration, car ownership and city shape were also relevant. Even if existent public transport modes in cities are not necessarily optimal, the classification models developed can give an insight for policy makers, in the sense that cities whose public transportation complexity cannot be explained by the models are more likely to have a suboptimal public transportation system.

Development and Validation of a Comprehensive Multivariate Dosimetric Model for Predicting Late Genitourinary Toxicity Following Prostate Cancer Stereotactic Body Radiotherapy

Purpose: Dosimetric predictors of toxicity after Stereotactic Body Radiation Therapy (SBRT) are not well-established. We sought to develop a multivariate model that predicts Common Terminology Criteria for Adverse Events (CTCAE) late grade 2 or greater genitourinary (GU) toxicity by interrogating the entire dose-volume histogram (DVH) from a large cohort of prostate cancer patients treated with SBRT on prospective trials. Methods: Three hundred and thirty-nine patients with late CTCAE toxicity data treated with prostate SBRT were identified and analyzed. All patients received 40 Gy in five fractions, every other day, using volumetric modulated arc therapy. For each patient, we examined 910 candidate dosimetric features including maximum dose, volumes of each organ [CTV, organs at risk (OARs)], V100%, and other granular volumetric/dosimetric indices at varying volumetric/dosimetric values from the entire DVH as well as ADT use to model and predict toxicity from SBRT. Training and validation subsets were generated with 90 and 10% of the patients in our cohort, respectively. Predictive accuracy was assessed by calculating the area under the receiver operating curve (AROC). Univariate analysis with student t-test was first performed on each candidate DVH feature. We subsequently performed advanced machine-learning multivariate analyses including classification and regression tree (CART), random forest, boosted tree, and multilayer neural network. Results: Median follow-up time was 32.3 months (range 3-98.9 months). Late grade ≥2 GU toxicity occurred in 20.1% of patients in our series. No single dosimetric parameter had an AROC for predicting late grade ≥2 GU toxicity on univariate analysis that exceeded 0.599. Optimized CART modestly improved prediction accuracy, with an AROC of 0.601, whereas other machine learning approaches did not improve upon univariate analyses. Conclusions: CART-based machine learning multivariate analyses drawing from 910 dosimetric features and ADT use modestly improves upon clinical prediction of late GU toxicity alone, yielding an AROC of 0.601. Biologic predictors may enhance predictive models for identifying patients at risk for late toxicity after SBRT.