重力再适应相关立位耐力评估与预测

Evaluation and Prediction of Orthostatic Tolerance Related to Gravity Re-adaptation

目的研发适宜于返回后立位耐力早期评估的安全检测方法,探索构建重力再适应相关立位耐力预测模型。方法 15名健康男性进行被动立位耐力测试,按原有评分标准,对被动立位测试数据中的3min、5 min、10 min数据分别计分;采用ROC法重新界定阈值进行分级,留一交叉验证,与原分级结果比对,验证一致性。利用15名健康男性头低位卧床模拟失重数据,基于贝叶斯原理,构建立位耐力预测模型,留一交叉验证后,选取4例飞行数据进一步验证。结果 3 min测试的ROC曲线下面积为0.944,5min和10 min的均为1.00,与原立位耐力测试分级结果一致性好。基于头低位模拟失重数据确立了5个立位耐力预测特征量及其似然比,留一交叉验证结果显示,ROC曲线下面积为0.75;4名参加飞行航天员验证结果显示,ROC曲线下面积为1,重力再适应立位耐力得到有效预测。结论采用新的方法对被动立位耐力3 min、5 min、10 min测试数据进行评估与20 min分级结果一致性好,具有应用前景。建立的重力再适应相关立位耐力预测模型具有较好预测能力。

Objective To develop a safe method for early evaluation of orthostatic tolerance soon after re-exposure to gravity and to construct a prediction model for orthostatic tolerance during gravity re-adaption. Methods Fifteen male volunteers performed 20 min passive orthostatic tolerance test. The results were rescored based on the 3 min,5 min and 10 min data with original standard. Threshold was redefined by ROC method,and orthostatic tolerance grade was reclassified. Leave one out cross validation was used to confirm the consistency of the reclassification results with the original one. Prediction model was constructed on head down bed rest test data of 15 male volunteers based on Bayes principle. After leave one out cross validation,the prediction model was further verified by spaceflight data. Results The area under the ROC curve for the 3 min test was 0. 944,the area of 5 min and 10 min were both 1. 00. Five prediction characteristic variables and its likelihood ratio were established based on the bed rest test data. The result of leave one out cross validation showed that the area under the ROC curve was 0. 75. The result of 4 spaceflight data showed that the area under the ROC curve was 1,and the orthostatic tolerance during re-adaptation gravity was effectively predicted. Conclusions The grading results of 3 min,5 min,and 10 min passive orthostatic tolerance test with new evaluation method was consistent with the 20 min original one. The orthostatic tolerance prediction model showed a good prediction ability.