The AIR-CPR Study: AI-Guided Chest Compressions

Official title

Utilizing Artificial Intelligence to Optimize Chest Compression Region During Cardio-pulmonary Resuscitation for Patients With Out-of-hospital Cardiac Arrest.

Overview

The AIR-CPR project aims to improve survival rates for patients with Out-of-Hospital Cardiac Arrest (OHCA) by utilizing Artificial Intelligence (AI) to optimize chest compression locations. Current guidelines recommend a standardized compression point (the lower half of the sternum), yet recent research indicates that this position can compress the aortic valve in approximately 48.7% of patients, significantly reducing the chances of successful resuscitation. This study will develop a deep learning model based on YOLO v8 to analyze real-time arterial pressure waveforms to identify proper aortic valve opening and closing. By identifying specific waveform features that humans cannot easily distinguish, the AI will guide rescuers to adjust the compression site-typically toward the left ventricle-to ensure optimal blood output. The project seeks to transform CPR from a standardized "one-size-fits-all" approach into a personalized, precision medicine intervention.

Detailed description

This three-year prospective study is designed to develop and clinically validate an "AI-Enhanced Arterial Waveform Monitor" to guide precision CPR.

1. Research Hypothesis and Objectives The study tests the hypothesis that AI can accurately predict aortic valve compression (confirmed by Transesophageal Echocardiography, TEE) by analyzing arterial pressure waveforms, thereby allowing rescuers to find the optimal compression site that avoids the aortic valve and maximizes cardiac output. 2. Implementation Phases

The project is divided into five distinct stages:

Case Preparation: Enrollment of 150 OHCA patients to collect synchronized TEE video and arterial pressure data.

Arterial Waveform Detection Model: Development of an algorithm to automatically segment continuous pressure signals into single-compression waveform samples.

Compression Region Detection Model: Training a YOLO v8-based model integrated with patient physiological data (age, sex, medical history) to distinguish between "compressed" and "non-compressed" aortic valve states.

Clinical External Testing: Enrolling an additional 75 patients to verify model accuracy against TEE "gold standard" findings.

Feasibility Assessment: Deploying the model as a "Resuscitation Support App" in 30 real-world clinical cases to evaluate its real-time guidance capability, speed, and impact on patient outcomes. 3. Technical Methodology

Data Extraction: Using binarization and interpolation curve fitting to extract high-quality numerical data directly from physiological monitor screens.

AI Architecture: Utilizing an improved YOLO v8 framework combined with an Attention-based architecture and Fully-connected neural networks to incorporate complex patient heterogeneities.

Clinical Intervention: When the AI identifies aortic valve compression, rescuers will be prompted to adjust the compression location (typically downward and to the left) until the valve is no longer obstructed. 4. Outcome Measures The study will evaluate the Identification Success Rate (AI vs. TEE), Avoidance Success Rate (successful repositioning), and traditional resuscitation metrics including ROSC, survival to discharge, and favorable neurologic outcomes.

Interventions

• Device Device: AI-Enhanced Arterial Waveform Monitor (AIR-CPR App)
  A deep learning application based on the YOLO v8 architecture that analyzes real-time arterial pressure waveforms from a femoral A-line. It identifies whether the current chest compression location is causing aortic valve compression (as confirmed by TEE) and provides immediate feedback to the resuscitation team.
• Procedure AI-Guided Chest Compression Repositioning
  When the AI application indicates aortic valve compression, the rescuer adjusts the mechanical chest compression (LUCAS) position. Based on literature and AI feedback, the adjustment typically involves moving the compression point downward and toward the left parasternal line to avoid the aortic valve and optimize left ventricular output.
• Diagnostic test Transesophageal Echocardiography (TEE)
  Used as the "Gold Standard" throughout the study. TEE is performed during CPR to record the actual opening and closing of the aortic valve and the deformation of cardiac chambers, providing the labels for AI training and the verification for clinical testing.

Primary outcome measures

• AI Identification Accuracy of Aortic Valve Compression [Time frame: Collected during the clinical testing phase and feasibility assessment (Years 2 and 3).]

Secondary outcome measures (5)

• Successful Avoidance of Aortic Valve Compression [Time frame: During the clinical feasibility assessment (Year 3).]
• Time Consumed for Compression Adjustment [Time frame: During the clinical feasibility assessment (Year 3).]
• Rate of Return of Spontaneous Circulation (ROSC) [Time frame: From the start of the emergency department resuscitation until hospital discharge or death (up to approximately 30 days).]
• Favorable Neurologic Outcome at Discharge [Time frame: At the time of hospital discharge (up to approximately 30 days).]
• Chest Compression Fraction (CCF) [Time frame: During the clinical feasibility assessment (Year 3).]

Eligibility criteria

Inclusion criteria

• Adults aged 20 years or older.
• Patients with out-of-hospital cardiac arrest (OHCA) undergoing 3.cardiopulmonary resuscitation (CPR) in the emergency department.

Cardiac arrest caused by non-traumatic factors.

Exclusion criteria

• Pregnant patients.
• Patients with obvious signs of death.
• Patients with a signed "Do Not Resuscitate" (DNR) order.
• Patients requiring extracorporeal cardio-pulmonary resuscitation (ECPR).
• Patients requiring Resuscitative Endovascular Balloon Occlusion of the Aorta (REBOA).
• Cardiac arrest caused by massive hemorrhage, aortic emergencies, tension pneumothorax, cardiac tamponade, or pulmonary embolism.
• History of severe aortic valve disease or previous aortic valve surgery.
• Patients for whom TEE or femoral arterial catheterization is contraindicated.
• Situations where the medical team is unable to perform TEE or femoral arterial catheterization during CPR.

Criteria are shown verbatim from the registry (in English). Final eligibility is always assessed by the study center.

Healthy volunteers: No

Study design

Observational model

Cohort

Study locations

Taiwan · 1 center

• Far Eastern Memorinal Hospital — New Taipei City

Publications

• Miao X, Wu Y, Wang J, Gao Y, Mao X, Yin J. Generative Semi-supervised Learning for Multivariate Time Series Imputation. Proceedings of the AAAI Conference on Artificial Intelligence 2021;35(10):8983-8991. DOI: 10.1609/aaai.v35i10.17086.
• Salloum R, Kuo CCJ. ECG-based biometrics using recurrent neural networks. 2017 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)2017:2062-2066.
• Mitra A, Kundu P, Gupta R. Hyperparameter Optimization of AutoRegressive Integrated Moving Average (ARIMA) Model-based Synthesis of Electrocardiogram. 2022 International Conference for Advancement in Technology (ICONAT)2022:1-6.
• Emergency Research. Content current as of: 09/21/2015 https://www.fda.gov/science-research/clinical-trials-and-human-subject-protection/protection-human-subjects-informed-consent-and-waiver-informed-consent-requirements-certain.
• Karhade J, Dash S, Ghosh SK, Dash DK, Tripathy RK. Time-Frequency-Domain Deep Learning Framework for the Automated Detection of Heart Valve Disorders Using PCG Signals. IEEE Transactions on Instrumentation and Measurement 2022;71:1-11. DOI: 10.1109/TIM.2022.3163156.
• Zhu Z, Wang H, Zhao T, et al. Classification of Cardiac Abnormalities From ECG Signals Using SE-ResNet. 2020 Computing in Cardiology2020:1-4.
• Nagasawa T, Iuchi K, Takahashi R, et al. Blood Pressure Estimation by Photoplethysmogram Decomposition into Hyperbolic Secant Waves. Applied Sciences 2022;12(4):1798. (https://www.mdpi.com/2076-3417/12/4/1798).
• Chirinos JA, Akers SR, Vierendeels JA, Segers P. A Unified Mechanism for the Water Hammer Pulse and Pulsus Bisferiens in Severe Aortic Regurgitation: Insights from Wave Intensity Analysis. Artery Res. 2018 Mar;21:9-12. doi: 10.1016/j.artres.2017.12.002. Epub 2017 Dec 21. No abstract available. PMID 29576810

Identifiers

NCT: NCT07431710 · 113046-F

Primary sources (government registries)

• ClinicalTrials.gov (US NIH)

View this study on ClinicalTrials.gov ↗