Division of Neurosurgery, University Surgical Centre, National University Hospital, Singapore.
Department of Electrical and Computer Engineering, National University of Singapore.
J Stroke Cerebrovasc Dis. 2022 Feb;31(2):106234. doi: 10.1016/j.jstrokecerebrovasdis.2021.106234. Epub 2021 Dec 10.
This study aims to develop and compare the use of deep neural networks (DNN) and support vector machines (SVM) to clinical prognostic scores for prognosticating 30-day mortality and 90-day poor functional outcome (PFO) in spontaneous intracerebral haemorrhage (SICH).
We conducted a retrospective cohort study of 297 SICH patients between December 2014 and May 2016. Clinical data was collected from electronic medical records using standardized data collection forms. The machine learning workflow included imputation of missing data, dimensionality reduction, imbalanced-class correction, and evaluation using cross-validation and comparison of accuracy against clinical prognostic scores.
32 (11%) patients had 30-day mortality while 177 (63%) patients had 90-day PFO. For prognosticating 30-day mortality, the class-balanced accuracies for DNN (0.875; 95% CI 0.800-0.950; McNemar's p-value 1.000) and SVM (0.848; 95% CI 0.767-0.930; McNemar's p-value 0.791) were comparable to that of the original ICH score (0.833; 95% CI 0.748-0.918). The c-statistics for DNN (0.895; DeLong's p-value 0.715), and SVM (0.900; DeLong's p-value 0.619), though greater than that of the original ICH score (0.862), were not significantly different. For prognosticating 90-day PFO, the class-balanced accuracies for DNN (0.853; 95% CI 0.772-0.934; McNemar's p-value 0.003) and SVM (0.860; 95% CI 0.781-0.939; McNemar's p-value 0.004) were better than that of the ICH-Grading Scale (0.706; 95% CI 0.600-0.812). The c-statistic for SVM (0.883; DeLong's p-value 0.022) was significantly greater than that of the ICH-Grading Scale (0.778), while the c-statistic for DNN was 0.864 (DeLong's p-value 0.055).
We showed that the SVM model performs significantly better than clinical prognostic scores in predicting 90-day PFO in SICH.
本研究旨在开发并比较深度学习神经网络（DNN）和支持向量机（SVM）在预测自发性脑出血（SICH）患者 30 天死亡率和 90 天不良功能结局（PFO）方面的应用，以临床预后评分作为参照。
我们对 2014 年 12 月至 2016 年 5 月期间的 297 例 SICH 患者进行了回顾性队列研究。使用标准化数据采集表从电子病历中收集临床数据。机器学习工作流程包括缺失数据的插补、降维、不平衡类别的校正，以及使用交叉验证进行评估，并与临床预后评分进行准确性比较。
32（11%）例患者在 30 天内死亡，177（63%）例患者在 90 天内出现不良功能结局。在预测 30 天死亡率方面，DNN（0.875；95%置信区间 0.800-0.950；McNemar 检验 p 值为 1.000）和 SVM（0.848；95%置信区间 0.767-0.930；McNemar 检验 p 值为 0.791）的平衡准确率与原始 ICH 评分（0.833；95%置信区间 0.748-0.918）相当。DNN（0.895；DeLong 检验 p 值为 0.715）和 SVM（0.900；DeLong 检验 p 值为 0.619）的 C 统计量虽然大于原始 ICH 评分（0.862），但差异无统计学意义。在预测 90 天 PFO 方面，DNN（0.853；95%置信区间 0.772-0.934；McNemar 检验 p 值为 0.003）和 SVM（0.860；95%置信区间 0.781-0.939；McNemar 检验 p 值为 0.004）的平衡准确率优于 ICH 分级量表（0.706；95%置信区间 0.600-0.812）。SVM 的 C 统计量（0.883；DeLong 检验 p 值为 0.022）明显大于 ICH 分级量表（0.778），而 DNN 的 C 统计量为 0.864（DeLong 检验 p 值为 0.055）。
我们发现 SVM 模型在预测 SICH 患者 90 天 PFO 方面的表现明显优于临床预后评分。
