Tissue dosimetry expansion and cross-validation of rat and mouse physiologically based pharmacokinetic models for trichloroethylene.

Trichloroethylene (TCE), a volatile liquid used as a degreasing agent, is a common environmental pollutant. In 2001, the EPA published a draft risk assessment for TCE that incorporates dosimetry predictions of physiologically based pharmacokinetic (PBPK) models. The current modeling effort represents an expansion and extensive tissue dosimetry validation of rodent PBPK models for TCE. The pharmacokinetics of TCE in male Sprague-Dawley (S-D) rats were characterized (1) during and after inhalation exposure to 50 or 500 ppm TCE, (2) following administration of 8 mg/kg TCE PO, and (3) following intra-arterial injection of 8 mg/kg TCE. Blood and tissues (including liver, kidney, fat, skeletal muscle, heart, spleen, gastrointestinal tract, and brain) were collected at selected time-points from 5 min up to 24 h post initial exposure. The fat compartment was modified to be diffusion-limited to predict the observed slow release of TCE from the fat. The addition of a deep liver compartment was necessary to accurately predict the slower hepatic clearance of TCE for all three exposure routes. Simulations of liver concentrations following gavage of male B6C3F1 mice with 300-2000 mg/kg TCE were also improved with the addition of a deep liver compartment. Liver predictions were calibrated and validated using a cross-validation technique novel to PBPK modeling. Splitting of compartments did not significantly affect predictions of TCE concentrations in the liver, fat, or venous blood. This model expansion and validation increases both the utility and our confidence in the current use of rodent TCE PBPK models in human health risk assessment.     

在线近红外光谱监测桂枝茯苓胶囊流化床干燥过程水分的方法研究

目的采用在线近红外光谱(NIRS)技术,建立桂枝茯苓胶囊流化床干燥过程水分实时监测模型。方法通过NIRS漫反射探头采集16个生产批次共176个样本进行建模,优选移动窗口平滑法进行光谱预处理,采用间隔偏最小二乘法(si PLS)结合移动窗口偏最小二乘法(mw PLS)筛选特征变量为4 759.45～5 338.00 cm-1、5 503.84～6 101.67 cm-1、8 512.25～8 809.24cm-1,采用偏最小二乘(PLS)法建立水分含量多变量校正模型。结果水分预测的交叉验证均方根误差(RMSECV)为0.243%,性能偏差比(RPD)值为13.384,预测相对偏差(RSEP)为0.270%。以8个生产批次对在线监控方法的可靠性进行持续验证,结果 40个样本的相对预测误差均小于4.7%。干燥过程水分实时监测趋势图显示可准确判断干燥终点,终点样本水分含量位于控制限内。结论在线NIRS结合PLS建立的定量模型,可应用于生产规模桂枝茯苓胶囊流化床干燥过程水分含量在线监控且预测性能稳健、准确。     

Validation of a predictive model for asthma admission in children: how accurate is it for predicting admissions?

We studied 364 index presentations to the Emergency Department of a children's hospital with a diagnosis of asthma. The admission rate for this group of children was about 31%. We developed a parsimonious multiple logistic regression model to predict asthma hospital admission based on asthma severity indicators. We then evaluated the model's predictive ability using two methods of cross-validation, using the same sample that was used for the predictive model, and using data from a split sample. The logistic regression model had a predictive accuracy of 90% (95% confidence interval 85–95%). The sensitivity and specificity were 86% and 88%, respectively. Cross-validation models confirmed that the predictive ability of the model was stable. In studies with limited sample sizes, it is possible to validate a model without setting aside a split sample for cross-validation.     

Artificial intelligence model validation before its application in clinical diagnosis assistance

The process of selecting an artificial intelligence(AI)model to assist clinical diagnosis of a particular pathology and its validation tests is relevant since the values of accuracy,sensitivity and specificity may not reflect the behavior of the method in a real environment.Here,we provide helpful considerations to increase the success of using an AI model in clinical practice.     

基于SVM的尿液粒子识别算法研究

本文将支持向量机的算法引入到尿沉渣有形成分的分类问题上.在提取特征的基础上,采用交叉验证法和精度等高线图进行核函数及参数的选择.根据支持向量机和数据集特点,设计出由两级分类器集成的支持向量机多分类器.得到了相应的混淆矩阵.临床实验数据分类评测以及与神经网络方法比较结果表明,提出的算法具有一定的优势.     

Machine learning in small sample neuroimaging studies: Novel measures for schizophrenia analysis

Novel features derived from imaging and artificial intelligence systems are commonly coupled to construct computer-aided diagnosis (CAD) systems that are intended as clinical support tools or for investigation of complex biological patterns. This study used sulcal patterns from structural images of the brain as the basis for classifying patients with schizophrenia from unaffected controls. Statistical, machine learning and deep learning techniques were sequentially applied as a demonstration of how a CAD system might be comprehensively evaluated in the absence of prior empirical work or extant literature to guide development, and the availability of only small sample datasets. Sulcal features of the entire cerebral cortex were derived from 58 schizophrenia patients and 56 healthy controls. No similar CAD systems has been reported that uses sulcal features from the entire cortex. We considered all the stages in a CAD system workflow: preprocessing, feature selection and extraction, and classification. The explainable AI techniques Local Interpretable Model-agnostic Explanations and SHapley Additive exPlanations were applied to detect the relevance of features to classification. At each stage, alternatives were compared in terms of their performance in the context of a small sample. Differentiating sulcal patterns were located in temporal and precentral areas, as well as the collateral fissure. We also verified the benefits of applying dimensionality reduction techniques and validation methods, such as resubstitution with upper bound correction, to optimize performance.     

Comparison between the more recent techniques for smoothing and derivative assessment in biomechanics

When analysising and evaluating human motion, two strictly interconnected problems arise: the data smoothing and the determination of velocities and accelerations from displacement data. Differentiating procedures magnify the noise superimposed on the useful kinematic data. A smoothing procedure is thus required to reduce the measurement noise before the differentiation can be carried out. In the paper two techniques for derivative assessment are presented, tested and compared. One of these is the procedure known as one of the best automatic smoothing and differentiating techniques: generalised cross validatory spline smoothing and differentiation (GCVC). The other, which has recently been presented, features an automatic model-based bandwidth-selection procedure (LAMBDA). The procedures have been tested with signals presented by other authors and available in the literature, by test signals acquired using the ELITE motion analyser and by synthetic data. The results show better or similar performance of LAMBDA compared with GCVC. In the cases in which the natural conditions at the signal boundaries are not met GCVC gives bad results (especially on the third derivative) whereas LAMBDA is not affected at all. Moreover, analysis time is dramatically lower for LAMBDA.