PDF(Pubmed)
Abstract：
OBJECTIVE: In this research, a non-invasive intracranial pressure (nICP) optical sensor was developed and evaluated in a clinical pilot study. The technology relied on infrared light to probe brain tissue, using photodetectors to capture backscattered light modulated by vascular pulsations within the brain\'s vascular tissue. The underlying hypothesis was that changes in extramural arterial pressure could affect the morphology of recorded optical signals (photoplethysmograms, or PPGs), and analysing these signals with a custom algorithm could enable the non-invasive calculation of intracranial pressure (nICP).
METHODS: This pilot study was the first to evaluate the nICP probe alongside invasive ICP monitoring as a gold standard. nICP monitoring occurred in 40 patients undergoing invasive ICP monitoring, with data randomly split for machine learning. Quality PPG signals were extracted and analysed for time-based features. The study employed Bland-Altman analysis and ROC curve calculations to assess nICP accuracy compared to invasive ICP data.
RESULTS: Successful acquisition of cerebral PPG signals from traumatic brain injury (TBI) patients allowed for the development of a bagging tree model to estimate nICP non-invasively. The nICP estimation exhibited 95% limits of agreement of 3.8 mmHg with minimal bias and a correlation of 0.8254 with invasive ICP monitoring. ROC curve analysis showed strong diagnostic capability with 80% sensitivity and 89% specificity.
CONCLUSIONS: The clinical evaluation of this innovative optical nICP sensor revealed its ability to estimate ICP non-invasively with acceptable and clinically useful accuracy. This breakthrough opens the door to further technological refinement and larger-scale clinical studies in the future.
BACKGROUND: NCT05632302, 11th November 2022, retrospectively registered.
METHODS: This pilot study was the first to evaluate the nICP probe alongside invasive ICP monitoring as a gold standard. nICP monitoring occurred in 40 patients undergoing invasive ICP monitoring, with data randomly split for machine learning. Quality PPG signals were extracted and analysed for time-based features. The study employed Bland-Altman analysis and ROC curve calculations to assess nICP accuracy compared to invasive ICP data.
RESULTS: Successful acquisition of cerebral PPG signals from traumatic brain injury (TBI) patients allowed for the development of a bagging tree model to estimate nICP non-invasively. The nICP estimation exhibited 95% limits of agreement of 3.8 mmHg with minimal bias and a correlation of 0.8254 with invasive ICP monitoring. ROC curve analysis showed strong diagnostic capability with 80% sensitivity and 89% specificity.
CONCLUSIONS: The clinical evaluation of this innovative optical nICP sensor revealed its ability to estimate ICP non-invasively with acceptable and clinically useful accuracy. This breakthrough opens the door to further technological refinement and larger-scale clinical studies in the future.
BACKGROUND: NCT05632302, 11th November 2022, retrospectively registered.
摘要：
目的：在这项研究中，我们开发了一种无创颅内压(nICP)光学传感器,并在一项临床试验研究中进行了评估.该技术依靠红外光探测脑组织，使用光电探测器捕获大脑血管组织内血管搏动调制的反向散射光。潜在的假设是，壁外动脉压的变化可能会影响记录的光信号的形态（光电体积描记图，或PPG)，使用自定义算法分析这些信号可以实现颅内压（nICP）的无创计算。
方法：这项初步研究首次将nICP探针与侵入性ICP监测作为金标准进行评估。NICP监测发生在接受有创ICP监测的40例患者中，数据随机拆分用于机器学习。提取并分析质量PPG信号的基于时间的特征。该研究采用Bland-Altman分析和ROC曲线计算来评估与侵入性ICP数据相比的nICP准确性。
结果：成功获得创伤性脑损伤（TBI）患者的脑PPG信号，可以开发一种套袋树模型来无创地估计nICP。nICP估计值显示出95％的一致性极限为3.8mmHg，偏差最小，与侵入性ICP监测的相关性为0.8254。ROC曲线分析具有较强的诊断能力,敏感性为80%,特异性为89%。
结论：对这种创新的光学nICP传感器的临床评估揭示了其能够以可接受的和临床有用的准确性无创估计ICP的能力。这一突破为未来进一步的技术改进和更大规模的临床研究打开了大门。
背景：NCT05632302，2022年11月11日，回顾性注册。
方法：这项初步研究首次将nICP探针与侵入性ICP监测作为金标准进行评估。NICP监测发生在接受有创ICP监测的40例患者中，数据随机拆分用于机器学习。提取并分析质量PPG信号的基于时间的特征。该研究采用Bland-Altman分析和ROC曲线计算来评估与侵入性ICP数据相比的nICP准确性。
结果：成功获得创伤性脑损伤（TBI）患者的脑PPG信号，可以开发一种套袋树模型来无创地估计nICP。nICP估计值显示出95％的一致性极限为3.8mmHg，偏差最小，与侵入性ICP监测的相关性为0.8254。ROC曲线分析具有较强的诊断能力,敏感性为80%,特异性为89%。
结论：对这种创新的光学nICP传感器的临床评估揭示了其能够以可接受的和临床有用的准确性无创估计ICP的能力。这一突破为未来进一步的技术改进和更大规模的临床研究打开了大门。
背景：NCT05632302，2022年11月11日，回顾性注册。
