AI Assistance for Clinicians Identifying Causes of Chest Pain From ECG Images

November 12, 2026 updated by: Shenda Hong, Peking University Health Science Center

A Two-Arm Parallel-Group Randomized Trial of ZERO-Assisted and Unassisted Clinicians Identifying Causes of Chest Pain From Real-World ECG Images

This study will assess whether reports generated by ZERO, a mobile system for interpreting electrocardiogram (ECG) images, improve clinicians' ability to identify causes of chest pain compared with interpretation without AI assistance.

A total of 120 licensed cardiology clinicians will be randomly assigned in a 1:1 ratio to ZERO-assisted or unassisted interpretation. Each clinician will assess 30 ECG cases during a single session lasting up to 60 minutes. The cases will include five examples from each of six diagnostic categories: unstable angina, non-ST-segment elevation myocardial infarction, ST-segment elevation myocardial infarction, aortic dissection, pulmonary embolism, and other diagnoses. No case or patient will be repeated within or across clinicians.

For each case, clinicians will first review the ECG and submit an initial diagnosis that cannot subsequently be changed. Clinicians in the ZERO-assisted group will then view a fixed report generated before clinician randomization. Clinicians in the unassisted group will review the ECG again without an AI report. All clinicians will then submit a final diagnosis.

Diagnostic performance will be assessed against clinical reference diagnoses established before the interpretation task and withheld from participating clinicians. The study will use ECG cases from completed clinical encounters and will not affect patient care. The hypothesis is that ZERO-assisted clinicians will achieve higher final diagnostic accuracy than unassisted clinicians.

Study Overview

Status

Not yet recruiting

Conditions

Intervention / Treatment

Study Type

Interventional

Enrollment (Estimated)

120

Phase

  • Not Applicable

Contacts and Locations

This section provides the contact details for those conducting the study, and information on where this study is being conducted.

Study Contact

Participation Criteria

Researchers look for people who fit a certain description, called eligibility criteria. Some examples of these criteria are a person's general health condition or prior treatments.

Eligibility Criteria

Ages Eligible for Study

18 years to 110 years (Adult, Older Adult)

Accepts Healthy Volunteers

No

Genders Eligible for Study

All

Description

Inclusion Criteria:

  • Licensed cardiology clinicians who routinely participate in the diagnosis or management of chest pain or cardiovascular disease and interpret electrocardiograms (ECGs).
  • Able to read Chinese-language reports and use the study platform.
  • Willing to provide electronic informed consent and sign a confidentiality statement.
  • Willing to complete all 30 assigned ECG interpretation tasks within the specified study period.

Exclusion Criteria:

  • Access to the reference diagnoses of the study cases, or involvement in screening, selecting, or adjudicating these cases.
  • Prior knowledge of the answers to the study cases through participation in model development or previous experiments.
  • Duplicate enrollment.
  • Inability to complete training in the use of the study platform.

Study Plan

This section provides details of the study plan, including how the study is designed and what the study is measuring.

How is the study designed?

Design Details

  • Primary Purpose: Diagnostic
  • Allocation: Randomized
  • Interventional Model: Parallel Assignment
  • Masking: None (Open Label)

Arms and Interventions

Participant Group / Arm
Intervention / Treatment
No Intervention: No AI Assistance

Clinicians will interpret 30 unique ECG cases, with five cases from each diagnostic category, during a single session lasting up to 60 minutes. For each case, clinicians will review the ECG image and submit an initial first-choice diagnosis, which will then be locked. They will subsequently review the same ECG image again without an AI report and submit a final first-choice diagnosis, which will also be locked.

Experimental: ZERO Assistance

Clinicians will first review each assigned ECG image and submit and lock an initial first-choice diagnosis. They will then view the fixed ZERO-generated report and submit and lock a final first-choice diagnosis. The report will include risk alerts for STEMI, aortic dissection, and pulmonary embolism, ranked diagnostic candidates, supporting explanations and available sources, and suggestions for further evaluation. If the report is unavailable or invalid, the failure will be recorded and the clinician may still submit a final diagnosis. Each clinician will complete 30 unique cases, with five cases per diagnostic category.

ZERO is a mobile system that processes de-identified ECG images and generates structured diagnostic reports. It integrates image preprocessing, a dedicated prediction model, risk stratification, knowledge retrieval, and report generation using GLM-4.6. Reports will be generated and stored before clinician randomization using a prespecified, frozen model, prompts, and retrieval configuration. During the interpretation task, the corresponding report will be displayed only after the clinician has submitted and locked the initial diagnosis. No additional patient information or clinician responses will be supplied to the system, and reports will not be modified during the task. No online adaptation will occur during the study.

Other Names:
  • ZERO

What is the study measuring?

Primary Outcome Measures

Outcome Measure
Measure Description
Time Frame
Finalfirst-choice diagnostic accuracy
Time Frame: 60 Minutes

For each clinician, final first-choice diagnostic accuracy will be calculated as the number of cases in which the final first-choice diagnosis matches the prespecified clinical reference diagnosis, divided by 30 and multiplied by 100%. Each clinician will assess five cases from each of six diagnostic categories. The outcome will be summarized as the mean clinician-level accuracy in each randomized group. Only clinicians with both an initial and a final diagnosis recorded for all 30 cases will be included in the diagnostic outcome analysis. The primary comparison will be between the ZERO-assisted and unassisted groups.

60 Minutes

Secondary Outcome Measures

Outcome Measure
Measure Description
Time Frame
Final Missed-Diagnosis Percentage for ST-Segment Elevation Myocardial Infarction
Time Frame: 60 Minutes

Among clinicians with complete initial and final diagnoses for all 30 cases, the percentage of cases with a reference diagnosis of ST-segment elevation myocardial infarction for which the final first-choice diagnosis is not ST-segment elevation myocardial infarction. Each clinician's percentage will be calculated as the number of such cases divided by five and multiplied by 100%. The outcome will be summarized as the mean clinician-level percentage in each randomized group. Lower values indicate fewer missed diagnoses in the ECG interpretation task.

60 Minutes
Final Missed-Diagnosis Percentage for Aortic Dissection
Time Frame: 60 Minutes

Among clinicians with complete initial and final diagnoses for all 30 cases, the percentage of cases with a reference diagnosis of aortic dissection for which the final first-choice diagnosis is not aortic dissection. Each clinician's percentage will be calculated as the number of such cases divided by five and multiplied by 100%. The outcome will be summarized as the mean clinician-level percentage in each randomized group. Lower values indicate fewer missed diagnoses in the ECG interpretation task.

60 Minutes
Final Missed-Diagnosis Percentage for Pulmonary Embolism
Time Frame: 60 Minutes

Among clinicians with complete initial and final diagnoses for all 30 cases, the percentage of cases with a reference diagnosis of pulmonary embolism for which the final first-choice diagnosis is not pulmonary embolism. Each clinician's percentage will be calculated as the number of such cases divided by five and multiplied by 100%. The outcome will be summarized as the mean clinician-level percentage in each randomized group. Lower values indicate fewer missed diagnoses in the ECG interpretation task.

60 Minutes
Proportion of Cases Revised From an Incorrect to a Correct Diagnosis
Time Frame: 60 Minutes

For each clinician, the proportion will be calculated as the number of cases in which the initial first-choice diagnosis does not match the reference diagnosis but the final first-choice diagnosis does match it, divided by 30 and multiplied by 100%. The outcome will be summarized as the mean clinician-level proportion in each randomized group. The analysis will include only clinicians with complete initial and final diagnoses for all 30 assigned cases. Higher values indicate that more cases were corrected during the second assessment.

60 Minutes
Proportion of Cases Revised From a Correct to an Incorrect Diagnosis
Time Frame: 60 Minutes

For each clinician, the proportion will be calculated as the number of cases in which the initial first-choice diagnosis matches the reference diagnosis but the final first-choice diagnosis does not match it, divided by 30 and multiplied by 100%. The outcome will be summarized as the mean clinician-level proportion in each randomized group. The analysis will include only clinicians with complete initial and final diagnoses for all 30 assigned cases. Lower values indicate that fewer initially correct diagnoses became incorrect during the second assessment.

60 Minutes

Collaborators and Investigators

This is where you will find people and organizations involved with this study.

Sponsor

Study record dates

These dates track the progress of study record and summary results submissions to ClinicalTrials.gov. Study records and reported results are reviewed by the National Library of Medicine (NLM) to make sure they meet specific quality control standards before being posted on the public website.

Study Major Dates

Study Start (Estimated)

November 19, 2026

Primary Completion (Estimated)

December 11, 2026

Study Completion (Estimated)

December 18, 2026

Study Registration Dates

First Submitted

November 12, 2026

First Submitted That Met QC Criteria

November 12, 2026

First Posted (Actual)

November 18, 2026

Study Record Updates

Last Update Posted (Actual)

November 18, 2026

Last Update Submitted That Met QC Criteria

November 12, 2026

Last Verified

November 1, 2026

More Information

Terms related to this study

Drug and device information, study documents

Studies a U.S. FDA-regulated drug product

No

Studies a U.S. FDA-regulated device product

No

This information was retrieved directly from the website clinicaltrials.gov without any changes. If you have any requests to change, remove or update your study details, please contact register@clinicaltrials.gov. As soon as a change is implemented on clinicaltrials.gov, this will be updated automatically on our website as well.