Investigating the interpretability of fetal status assessment using antepartum cardiotocographic records

Background Cardiotocography (CTG) interpretation plays a critical role in prenatal fetal monitoring. However, the interpretation of fetal status assessment using CTG is mainly confined to clinical research. To the best of our knowledge, there is no study on data analysis of CTG records to explore the causal relationships between the important CTG features and fetal status evaluation. Methods For analyses, 2126 cardiotocograms were automatically processed and the respective diagnostic features measured by the Sisporto program. In this paper, we aim to explore the causal relationships between the important CTG features and fetal status evaluation. First, we utilized data visualization and Spearman correlation analysis to explore the relationship among CTG features and their importance on fetal status assessment. Second, we proposed a forward-stepwise-selection association rule analysis (ARA) to supplement the fetal status assessment rules based on sparse pathological cases. Third, we established structural equation models (SEMs) to investigate the latent causal factors and their causal coefficients to fetal status assessment. Results Data visualization and the Spearman correlation analysis found that thirteen CTG features were relevant to the fetal state evaluation. The forward-stepwise-selection ARA further validated and complemented the CTG interpretation rules in the fetal monitoring guidelines. The measurement models validated the five latent variables, which were baseline category (BCat), variability category (VCat), acceleration category (ACat), deceleration category (DCat) and uterine contraction category (UCat) based on fetal monitoring knowledge and the above analyses. Furthermore, the interpretable models discovered the cause factors of fetal status assessment and their causal coefficients to fetal status assessment. For instance, VCat could predict BCat, and UCat could predict DCat as well. ACat, BCat and DCat directly affected fetal status assessment, where ACat was the important causal factor. Conclusions The analyses revealed the interpretation rules and discovered the causal factors and their causal coefficients for fetal status assessment. Moreover, the results are consistent with the computerized fetal monitoring and clinical knowledge. Our approaches are conducive to evidence-based medical research and realizing intelligent fetal monitoring.

Variable selection with missing data in both covariates and outcomes: Imputation and machine learning

Variable selection in the presence of both missing covariates and outcomes is an important statistical research topic. Parametric regression are susceptible to misspecification, and as a result are sub-optimal for variable selection. Flexible machine learning methods mitigate the reliance on the parametric assumptions, but do not provide as naturally defined variable importance measure as the covariate effect native to parametric models. We investigate a general variable selection approach when both the covariates and outcomes can be missing at random and have general missing data patterns. This approach exploits the flexibility of machine learning models and bootstrap imputation, which is amenable to nonparametric methods in which the covariate effects are not directly available. We conduct expansive simulations investigating the practical operating characteristics of the proposed variable selection approach, when combined with four tree-based machine learning methods, extreme gradient boosting, random forests, Bayesian additive regression trees, and conditional random forests, and two commonly used parametric methods, lasso and backward stepwise selection. Numeric results suggest that, extreme gradient boosting and Bayesian additive regression trees have the overall best variable selection performance with respect to the [Formula: see text] score and Type I error, while the lasso and backward stepwise selection have subpar performance across various settings. There is no significant difference in the variable selection performance due to imputation methods. We further demonstrate the methods via a case study of risk factors for 3-year incidence of metabolic syndrome with data from the Study of Women’s Health Across the Nation.

Reliability of Preoperative Planning Method That Considers Latent Medial Joint Laxity in Medial Open-Wedge Proximal Tibial Osteotomy

Background: Soft tissue laxity around the knee joint has been recognized as a crucial factor affecting correction error during medial open-wedge proximal tibial osteotomy (MOWPTO). Medial laxity in particular, which represents the changes in joint-line convergence angle (JLCA), affects soft tissue correction. Purpose: The purpose of this study was to quantify medial laxity and develop a preoperative planning method that considers medial laxity. Study Design: Cohort study; Level of evidence, 3. Methods: This study retrospectively reviewed 139 knees in 117 patients who underwent navigation-assisted MOWPTO from January 2014 to July 2019 for symptomatic medial compartment osteoarthritis with varus alignment >5°. We compared the results of 2 preoperative planning methods: conventional Miniaci (n = 47) and latent medial laxity reduction (LMLR) (n = 92). We evaluated the incidence of undercorrection, acceptable correction, and overcorrection. The radiologic parameters were analyzed using multiple linear regression with a stepwise selection model to establish an equation for the optimal preoperative planning method. The intraclass correlation coefficients (ICCs) of intraobserver, interobserver, and intermethod reliability were calculated. Results: The Miniaci method showed a higher incidence of overcorrection (55.3%) than the LMLR method (22.8%) at postoperative 6 months ( P = .0006). Multiple linear regression with a stepwise selection model revealed a high correlation coefficient ( R 2 = 0.888) for the following equation: Adjusted planned correction angle = 0.596 + 0.891 × Target correction angle – 0.255 × Δ JLCA valgus. Upon simplification, the following equation showed the highest intermethod ICC value (0.991): Target correction angle – ⅓Δ JLCA valgus, while the Miniaci method showed a relatively low ICC value of 0.875. Conclusion: There was a risk of overcorrection after MOWPTO using the conventional Miniaci method. An equation that considers medial laxity may help during preoperative planning for optimal correction during MOWPTO.

Prognostic significance of serum cytokines during acute exacerbation of idiopathic interstitial pneumonias treated with thrombomodulin

BackgroundAcute exacerbation (AE) has been reported to herald a poor prognosis in idiopathic pulmonary fibrosis and is now thought to do so in idiopathic interstitial pneumonias (IIPs). However, the pathophysiology of AE-IIPs is not sufficiently understood. In our previously reported SETUP trial, we found better survival in patients with AE-IIPs treated with corticosteroids and thrombomodulin than in those treated with corticosteroids alone. In that study, we collected serum samples to evaluate changes in cytokine levels and retrospectively examined the prognostic significance and pathophysiological role of serum cytokines in patients with AE-IIPs.MethodsThis study included 28 patients from the SETUP trial for whom serial serum samples had been prospectively obtained. AE-IIPs were diagnosed using the Japanese Respiratory Society criteria. All patients were treated with intravenous thrombomodulin and corticosteroids from 2014 to 2016. Serum levels of 27 cytokines were measured using Bio-Plex. The high-resolution CT pattern at the time of diagnosis of AE was classified as diffuse or non-diffuse.ResultsUnivariate analysis revealed that higher serum levels of interleukin (IL)-2, IL-7, IL-9, IL-12, IL13, basic fibroblast growth factor, granulocyte-macrophage colony-stimulating factor, interferon-γ inducible protein-10, platelet-derived growth factor and regulated on activation, normal T cell expressed and secreted (RANTES) at AE were significant predictors of 90-day survival. The HRCT pattern was also a significant clinical predictor of 90-day survival. Multivariate analysis with stepwise selection identified a higher serum RANTES level at AE to be a significant predictor of 90-day survival, including after adjustment for HRCT pattern. Multivariate analysis with stepwise selection suggested that a marked increase in the serum IL-10 level on day 8 could predict 90-day mortality.ConclusionsA higher serum RANTES level at AE the time of diagnosis predicted a good survival outcome, and an elevated serum IL-10 level on day 8 predicted a poor survival outcome.Trial registration numberUMIN000014969.

Assessment of coal pillar stability using principal component analysis and stepwise selection and elimination

Prediction of pillar stability is one of the most critical tasks in underground mining industries. This pillar stability analysis requires many input parameters and some of them are difficult to be determined. Various statistical based analysis is presented in literature for assessing pillar stability successfully. In the present work, the data from three mines had been to determine the factor of safety. A total of 63 pillar cases had been collected from the mines. Principal component analysis (PCA) and Stepwise selection and elimination (SSE) models were developed by using multi variate linear regression (MLR) on 45 data sets and subsequently the proposed models were validated on 18 different data sets. The value of coefficient of determination (R2) is 0.86 and 0.84 for PCA and SSE respectively. The root mean square error for PCA and SSE are found to be 0.112 and 0.123 respectively. On validation of the proposed model developed by PCA and SSE, the PCA model provided a better validation results. Hence, PCA is recommended for modelling pillar stability.

Predicting Surgical Margins in Patients With Borderline Resectable and Locally Advanced Pancreatic Cancer Undergoing Resection

Background Many patients with borderline resectable/locally advanced pancreatic ductal adenocarcinoma (borderline resectable [BR]/locally advanced [LA] pancreatic ductal adenocarcinoma [PDAC]) undergoing resection will have positive resection margins (R1), which is associated with poor prognosis. It might be useful to preoperatively predict the margin (R) status. Methods Data from patients with BR/LA PDAC who underwent a pancreatectomy between 2008 and 2018 at Brigham and Women’s Hospital were retrospectively reviewed. Logistic regression analysis was used to evaluate the association between R status and relevant preoperative factors. Significant predictors of R1 resection on univariate analysis ( P < .1) were entered into a stepwise selection using the Akaike information criterion to define the final model. Results A total of 142 patients with BR/LA PDAC were included in the analysis, 60(42.3%) had R1 resections. In stepwise selection, the following factors were identified as positive predictors of an R1 resection: evidence of lymphadenopathy at diagnosis (OR = 2.06, 95% CI: 0.99-4.36, P = .056), the need for pancreaticoduodenectomy (OR = 3.81, 96% CI: 1.15-15.70, P = .040), extent of portal vein/superior mesenteric vein involvement at restaging (<180°, OR = 3.57, 95% CI: 1.00-17.00, P = .069, ≥180°, OR = 7,32, 95% CI: 1.75-39.87, P = .010), stable CA 19-9 serum levels (less than 50% decrease from diagnosis to restaging, OR = 2.27, 95% CI: 0.84-6.36 P = .107), and no preoperative FOLFIRINOX (OR = 2.17, 95% CI: 0.86-5.64, P = .103). The prognostic nomogram based on this model yielded a probability of achieving an R1 resection ranging from <5% (0 factors) to >70% (all 5 factors). Conclusions Relevant preoperative clinicopathological characteristics accurately predict positive resection margins in patients with BR/LA PDAC before resection. With further development, this model might be used to preoperatively guide surgical decision-making in patients with BR/LA PDAC.

The value of Bayesian predictive projection for variable selection: an example of selecting lifestyle predictors of young adult well-being

Background Variable selection is an important issue in many fields such as public health and psychology. Researchers often gather data on many variables of interest and then are faced with two challenging goals: building an accurate model with few predictors, and making probabilistic statements (inference) about this model. Unfortunately, it is currently difficult to attain these goals with the two most popular methods for variable selection methods: stepwise selection and LASSO. The aim of the present study was to demonstrate the use predictive projection feature selection – a novel Bayesian variable selection method that delivers both predictive power and inference. We apply predictive projection to a sample of New Zealand young adults, use it to build a compact model for predicting well-being, and compare it to other variable selection methods. Methods The sample consisted of 791 young adults (ages 18 to 25, 71.7% female) living in Dunedin, New Zealand who had taken part in the Daily Life Study in 2013–2014. Participants completed a 13-day online daily diary assessment of their well-being and a range of lifestyle variables (e.g., sleep, physical activity, diet variables). The participants’ diary data was averaged across days and analyzed cross-sectionally to identify predictors of average flourishing. Predictive projection was used to select as few predictors as necessary to approximate the predictive accuracy of a reference model with all 28 predictors. Predictive projection was also compared to other variable selection methods, including stepwise selection and LASSO. Results Three predictors were sufficient to approximate the predictions of the reference model: higher sleep quality, less trouble concentrating, and more servings of fruit. The performance of the projected submodel generalized well. Compared to other variable selection methods, predictive projection produced models with either matching or slightly worse performance; however, this performance was achieved with much fewer predictors. Conclusion Predictive projection was used to efficiently arrive at a compact model with good predictive accuracy. The predictors selected into the submodel – felt refreshed after waking up, had less trouble concentrating, and ate more servings of fruit – were all theoretically meaningful. Our findings showcase the utility of predictive projection in a practical variable selection problem.