基于影像组学参数评估376例幕上自发性脑出血患者的功能状态

目的 探讨基于首次平扫CT的影像组学列线图模型对幕上自发性脑出血患者90 d功能状态的预测价值。方法 回顾性分析幕上自发性脑出血(sICH)患者376例，根据90 d改良Rankin量表(mRS)评分结果将患者分为预后不良组121例(mRS评分为4～6分)和预后良好组255例(mRS评分为0～3分)。从首次平扫CT提取影像组学特征，计算影像组学评分(Rad-score)并构建影像组学模型。分析筛选临床因素用于构建临床模型，并结合Rad-score构建列线图模型。分析比较上述3种模型的预测效能。结果 最终筛选出20个特征用于构建影像组学模型，临床模型由年龄(OR=1.045,95%CI:1.023～1.066)、格拉斯哥昏迷评分GCS≤8(OR=4.128,95%CI:2.161～7.887)、血肿破入脑室(OR=3.071,95%CI:1.744～5.408)和血肿体积>30 mL(OR=5.802,95%CI:3.327～10.117)构成。在训练集，列线图模型的曲线下面积(0.892)高于临床模型(0.814)及影像组学模型(0.862),差异有统计学意义(Z值分别为3.35...

Functional outcomes of 376 patients with supratentorial spontaneous intracerebral hemorrhage based on radiomic parameters

Objective To investigate the value of a radiomic nomogram model based on initial CT scan to predict the 90-day functional outcomes of supratentorial spontaneous intracerebral hemorrhage(sICH). Methods Clinical data of 376 patients with supratentorial sICH were retrospectively analyzed. According to the 90-day modified Rankin Scale(mRS), the patients were divided into poor outcome group(n=121, mRS score 4-6)and good outcome group(n=255, mRS score 0-3). Radiomic features were extracted from the initial CT scan, radiomic score(Rad-score)was calculated and a radiomic model was constructed. The clinical risk factors of poor outcome were selected to construct a clinical model, and then combined with Rad-score to construct a nomogram model. The diagnostic values of the three models were compared. Results Altogether 20 features were selected to establish the radiomic model. The clinical model was composed of age(OR=1.045, 95%CI: 1.023-1.066), Glasgow Coma Scale(GCS)≤8(OR=4.128, 95%CI: 2.161-7.887), hematoma breaking into ventricle(OR=3.071, 95%CI: 1.744-5.408)and hematoma volume >30 mL(OR=5.802, 95%CI: 3.327-10.117). In the training set, the area under the curve(AUC)of the nomogram model was 0.892, which was higher than that of the clinical model(0.814)and the radiomics model(0.862), and the difference was statistically significant(Z=3.356, 2.231, P=0.000 8, 0.025 7). Conclusion The nomogram model can effectively predict the 90-day outcomes for supratentorial sICH patients, which is helpful for clinical decision-making.