人工智能驱动的床旁超声颈动脉可压缩性评估用于危重症及复苏后患者血压估算:一项前瞻性观察性研究
该前瞻性观察性研究纳入急诊科56例患者(30例危重症、26例复苏后自主循环恢复者),采集372段标准化床旁超声颈动脉压迫视频,用深度学习模型RealCAC-Net量化颈动脉可压缩性(CAC)。CAC与收缩压(r=-0.697)、平均动脉压(r=-0.656)、舒张压(r=-0.576)呈强负相关;识别收缩压<60 mmHg的AUC为0.90(95%CI 0.87–0.93),识别平均动脉压<40 mmHg的AUC为0.91(95%CI 0.88–0.94),回归模型连续估算血压的均方根误差最低为8.3 mmHg。模型在低血压范围内表现最佳。作者认为CAC可能为侵入性监测不可即时获得时提供无创血流动力学监测工具,但需进一步验证。
要点
- 在危重症及复苏后患者中,AI量化的颈动脉可压缩性与收缩压(r=-0.697)、平均动脉压(r=-0.656)、舒张压(r=-0.576)呈强负相关。
- 识别收缩压<60 mmHg的AUC为0.90(95%CI 0.87–0.93),识别平均动脉压<40 mmHg的AUC为0.91(95%CI 0.88–0.94)。
- 回归模型可连续估算血压,均方根误差最低为8.3 mmHg。
- 模型在低血压范围内表现最佳,作者认为CAC可能用于侵入性监测不可即时获得时的无创血流动力学监测。
- 作者指出这是概念验证研究,需进一步验证研究。
适用人群与场景
- 急诊科收治的危重症患者及复苏后自主循环恢复(ROSC)患者。
研究设计与关键结果
- 研究类型为前瞻性观察性研究,纳入56例患者(30例危重症、26例复苏后ROSC),分析372段床旁超声颈动脉压迫视频。
- CAC与收缩压(r=-0.697)、平均动脉压(r=-0.656)、舒张压(r=-0.576)呈强负相关。
- 识别收缩压<60 mmHg的AUC为0.90(95%CI 0.87–0.93),识别平均动脉压<40 mmHg的AUC为0.91(95%CI 0.88–0.94)。
- 回归模型连续估算血压的均方根误差最低为8.3 mmHg。
- 模型在低血压范围内表现最佳。
证据类型、等级与局限
- 证据类型:前瞻性观察性研究(概念验证研究)。
- 局限:作者指出这是概念验证研究,需进一步验证研究(Additional validation studies are warranted)。
对床旁实践的改变
- 作者认为CAC可能为侵入性监测不可即时获得时提供无创血流动力学监测工具,尤其在低血压范围内;这是概念验证研究,需进一步验证。
提炼说明
- 本条依据论文摘要整理;摘要未涉及「诊断标准与关键阈值」「处置流程与药物剂量」「红旗征与不能漏的情况」「指南或共识出处」,故未列出这些栏目;全文请点「阅读原文」。
英文原文(摘要)· 点击展开
BACKGROUND: Accurate, real-time blood pressure (BP) monitoring is critical in emergency and critical care, but current methods are limited. Invasive arterial catheterization, the gold standard, is often delayed in hypotensive patients, whereas noninvasive cuffs can be unreliable in a state of low perfusion. We hypothesized that carotid artery compressibility measured by point-of-care ultrasound (POCUS) and analyzed via artificial intelligence (AI) could be used to estimate arterial BP noninvasively. METHODS: We conducted a prospective observational study enrolling critically ill patients and post-return of spontaneous circulation (ROSC) patients in the emergency department. Standardized POCUS-guided carotid artery compression (POCUS-CAC) was performed. Video clips of POCUS-CAC were analyzed with RealCAC-Net, a deep learning model for quantifying carotid artery compressibility (CAC). The AI-derived maximum CAC value and concurrently measured BP were recorded. Diagnostic and regression analyses were conducted to evaluate the performance of CAC in classifying and predicting BP. RESULTS: A total of 372 ultrasound clips from 56 patients (30 critically ill, 26 post-ROSC) were analyzed. CAC demonstrated strong inverse correlations with systolic (r = −0.697), mean (r = −0.656), and diastolic (r = −0.576) arterial pressures. The model achieved excellent diagnostic performance for identifying hypotension: area under the curve of 0.90 (95% confidence interval (CI), 0.87–0.93) for systolic BP < 60 mmHg and 0.91 (95% CI, 0.88–0.94) for mean BP < 40 mmHg. Regression models enabled continuous BP estimation with root mean squared errors as low as 8.3 mmHg. The model performed best in hypotensive ranges. CONCLUSIONS: This proof-of-concept study demonstrated potential diagnostic utility of carotid artery compressibility for blood pressure estimation in critically ill and post-ROSC patients with hypotensive conditions. Our findings suggest that CAC may provide a noninvasive tool for hemodynamic monitoring, particularly in settings where invasive monitoring is not immediately available. Additional validation studies are warranted.
参考来源
- Maeng SJ, Park S, Jo IJ, et al. AI-driven carotid artery compressibility assessment via point-of-care ultrasound for blood pressure estimation in critically ill and post-resuscitation patients: a prospective observational study. Crit Care. 2025. PMID: 41316299. doi:10.1186/s13054-025-05772-2
供执业医师抢救与诊疗参考,不替代临床判断;剂量请结合体重、肝肾功能和本院药品规格复核,以指南原文为准。