The goal of this observational study is to evaluate the ability of artificial intelligence (AI) models to interpret arterial waveform analysis data obtained from a hemodynamic monitoring system in adult patients undergoing elective surgery. The main questions it aims to answer are:
Can AI models (ChatGPT-4 and Gemini 2.0) accurately detect hemodynamic abnormalities in arterial waveform data? How well do AI-generated diagnoses align with expert anesthesiologist assessments? Are AI-generated treatment recommendations clinically appropriate?
Participants will:
Undergo standard hemodynamic monitoring with an arterial waveform analysis device (MostCare).
Have their anonymized hemodynamic data analyzed by AI models for abnormality detection, diagnosis suggestions, and treatment recommendations.
Have AI-generated results reviewed and validated by experienced anesthesiologists.
This study aims to assess whether AI models can serve as decision-support tools in perioperative and critical care settings by improving the interpretation of complex hemodynamic data, potentially enhancing patient safety, diagnostic accuracy, and clinical efficiency.
Age ≥ 18 years
Undergoing elective surgery with arterial waveform monitoring as part of standard perioperative care
Hemodynamic data successfully recorded using the MostCare hemodynamic monitoring system
Able to provide informed consent to participate in the study
Incomplete or corrupted hemodynamic data (e.g., signal artifacts preventing reliable analysis)
Emergency surgery cases
Patients with severe arrhythmias or hemodynamic instability that might interfere with arterial waveform interpretation
Refusal to participate or withdrawal of consent
This study will include adult patients (≥18 years old) undergoing elective surgery requiring intraoperative arterial waveform monitoring as part of routine perioperative care. The population will consist of patients from two tertiary-level hospitals where advanced hemodynamic monitoring with the MostCare system is regularly utilized. Participants will be selected based on their eligibility for continuous arterial pressure monitoring, ensuring a standardized dataset for AI analysis. The study population will represent a diverse range of surgical procedures, including but not limited to: General surgery (e.g., abdominal, hepatobiliary, colorectal procedures)
Cohort
Prospective
predictions
Other interventionpredictions of learning language models
Accuracy
Accuracy of AI models in detecting hemodynamic abnormalities (True or False).
Concordance Between AI-Generated Diagnoses and Expert Anesthesiologist Diagnoses
AI-generated diagnostic suggestions will be compared with the final diagnosis made by anesthesiologists (True or False).
Clinical Appropriateness of AI-Generated Treatment Recommendations
The relevance and accuracy of AI-suggested treatments will be evaluated against standard clinical management (True or False).
Click on the lead sponsor to view all of their trials.
