23研赛 E题 出血性脑卒中临床智能诊疗建模

2744557306

为大家分享文章
文章的主要内容
搜集 2015 年 5 月至 2020 年 5 月于我院就诊的 232 例 sICH患者的临床和影像资料，以发病 24 小时内复查影像中血肿体积较首次血肿体积增加>6 mL 或 33%定义为发生早期 HE。按照初始 BAT 评分方法计算每位患者的评分，评价初始 BAT评分识别早期 HE 患者的性能。接着，将所有 sICH 患者以 7：3 比例随机分成训练集（n=162）和验证集（n=70）；在训练集中基于初始 BAT 评分的三个变量（混合征、低密度征和发病至首次 CT 检查时间），利用 5 种常用机器学习算法（随机森林算法、梯度提升树、朴素贝叶斯[Naive Bayes, NB]算法、逻辑回归算法和 k 近邻算法）构建修正的 BAT 评分，并在验证集中验证修正BAT 评分。绘制受试者工作特征曲线（receiver operating characteristic重庆医科大学博士研究生学位论文4curve, ROC）评价所有 BAT 评分的预测性能，采用 DeLong 检验来比较所有 BAT 评分的预测性能，采用决策曲线分析方法评价修正 BAT 评分的临床应用价值。
! C& b6 \0 V, t- L' d& J结果：
初始 BAT评分识别早期 HE 患者的 ROC 曲线下面积（areaunder the curve, AUC）为 0.57。在 5 个修正 BAT 评分中，基于 NB 算法的修正 BAT 评分预测早期HE 表现最佳，在训练集中其AUC 为0.83，在验证集中其 AUC 为 0.77。DeLong 检验结果表明，基于 NB 算法的修正 BAT 评分在训练集和验证集中的性能均显著优于初始 BAT 评分（P<0.001）。决策曲线分析结果显示，基于 NB 算法的修正 BAT 评分预测早期 HE 的临床适用性优于其他修正 BAT 评分
8 p7 C2 C# r/ N) L# C. \第一章：机器学习在改进自发性脑出血早期血肿扩大预测评分中的应用研究
第二章： CT 影像组学在自发性脑出血早期血肿扩大预测中的应用研究
第三章： CT 影像组学在自发性脑出血不良功能预后评估中的应用研究
$ G2 [% `" n2 t9 [第四章：基于 CT 平扫的影像学特征评估自发性脑出血预后的研究进展
第二篇文章
主要内容
Intracerebral hemorrhage (ICH) has one of the worst prognoses among patients with stroke.Surgical measures have been adopted to relieve the mass effect of the hematoma, anddeveloping targeted therapy against secondary brain injury (SBI) after ICH is equallyessential. Numerous preclinical and clinical studies have demonstrated thatperihematomal edema (PHE) is a quantifiable marker of SBI after ICH and is associatedwith a poor prognosis. Thus, PHE has been considered a promising therapeutic target forICH. However, the findings derived from existing studies on PHE are disparate and unclear.Therefore, it is necessary to classify, compare, and summarize the existing studies on PHE.In this review, we describe the growth characteristics and relevant underlying mechanism ofPHE, analyze the contributions of different risk factors to PHE, present the potential impactof PHE on patient outcomes, and discuss the currently available therapeutic strategies.
  [2 W6 _4 w& X8 x' Y第三篇文章
We attempt to generate a definition of delayed perihematomal edema expansion (DPE)and analyze its time course, risk factors, and clinical outcomes. A multi-cohort data wasderived from the Chinese Intracranial Hemorrhage Image Database (CICHID). A noncontrast computed tomography (NCCT) -based deep learning model was constructed forfully automated segmentation hematoma and perihematomal edema (PHE). Time courseof hematoma and PHE evolution correlated to initial hematoma volume was volumetricallyassessed. Predictive values for DPE were calculated through receiver operatingcharacteristic curve analysis and were tested in an independent cohort. Logisticregression analysis was utilized to identify risk factors for DPE formation and pooroutcomes. The test cohort’s Dice scores of lesion segmentation were 0.877 and 0.642for hematoma and PHE, respectively. Overall, 1201 patients were enrolled for time-courseanalysis of ICH evolution. A total of 312 patients were further selected for DPE analysis.Time course analysis showed the growth peak of PHE approximately concentrates in 14days after onset. The best cutoff for DPE to predict poor outcome was 3.34 mL ofabsolute PHE expansion from 4-7 days to 8-14 days (AUC=0.784, sensitivity=72.2%,specificity=81.2%), and 3.78 mL of absolute PHE expansion from 8-14 days to 15-21days (AUC=0.682, sensitivity=59.3%, specificity=92.1%) in the derivation sample.Patients with DPE was associated with worse outcome (OR: 12.340, 95%CI: 6.378-23.873, P<0.01), and the larger initial hematoma volume (OR: 1.021, 95%CI: 1.000-1.043, P=0.049) was the significant risk factor for DPE formation. This study constructed awell-performance deep learning model for automatic segmentations of hematoma andPHE. A new definition of DPE was generated and is confirmed to be related to poor

2 C: o2 u( |5 P
  X. A' Y( z: H  M8 r; K% W4 v