AI-Based Quality Control for Radiology Reporting

NCT07865741 | Not Yet Recruiting

• 1 other identifier: TJ-IRB202609054
• Study Type: interventional
• Target: 109
• Locations: 0 countries
• Sites: N/A

Brief Summary

The goal of this clinical trial is to learn whether an artificial intelligence (AI)-based radiology report quality-control system can improve the accuracy and workflow efficiency of routine radiology reporting among active radiologists at Tongji Hospital. The main questions it aims to answer are:

• Does access to the AI quality-control system reduce quality problems remaining in final radiology reports?
• Does access to the AI quality-control system improve reporting and reviewing efficiency?
• In the two-stage reporting workflow, do the effects of AI differ depending on whether AI is available to the reporting radiologist, the reviewing radiologist, or both? Researchers will compare radiologists assigned to receive access to Radiology Report GPT with radiologists who continue the usual reporting workflow without access to the study AI system. Radiologists are randomized at the individual level. Because CT and MR reports commonly involve both a reporting radiologist and a reviewing radiologist, report-level analyses will classify reports according to AI availability at the two workflow stages, resulting in four combinations of reporter and reviewer AI exposure. The primary randomized evaluation will cover the first month after intervention initiation, with prespecified secondary and exploratory analyses extending through 3 months. Participants will:
• Be assigned to either the AI quality-control group or the usual-workflow control group.
• Continue their routine radiology reporting and reviewing work during the study period.
• If assigned to the AI group, receive access to Radiology Report GPT, which automatically pre-reviews submitted radiology reports and provides quality-control alerts and suggested revisions. Radiologists will retain final decision authority over all reports. Final-report accuracy will be assessed independently of the intervention AI system and will include human review. Additional prespecified outcomes will examine report revisions, AI interaction measures, stage-specific and cross-stage workflow effects, recommended imaging follow-up and other downstream health-care utilization, patient satisfaction, and referring-clinician satisfaction.

Conditions

Radiology; Medical Errors

Outcome Measures

Primary Outcomes (2)

• Presence of at Least One Confirmed Residual Quality-Control Problem in the Final Radiology Report

  For each eligible final radiology report, this binary outcome indicates whether at least one confirmed prespecified residual quality-control problem is present (yes/no). The outcome-assessment process will be independent of the intervention AI system and will include human review. Prespecified categories include textual/terminology errors, sex or examination-type inconsistencies, formatting/numerical errors, findings-impression inconsistencies, lesion location/count discrepancies, omissions, and missing or incorrect diagnostic classifications.

  During the first month after intervention initiation

• Radiology Report Turnaround Time

  Turnaround time will be measured for each eligible radiology report as the elapsed time, from completion of the imaging examination to final report approval. Both timestamps will be obtained from the hospital information systems. Lower values indicate faster completion of the radiology reporting workflow.

  During the first month after intervention initiation

Secondary Outcomes (2)

• Inter-Report Interval

  During the first month after intervention initiation

• Inter-Check Interval

  During the first month after intervention initiation

Other Outcomes (3)

• Patient Adherence

  During the 3-month secondary evaluation period

• Patient Satisfaction

  During the 3-month secondary evaluation period

• Referring-Clinician Satisfaction

  During the 3-month secondary evaluation period

Study Arms (2)

AI Quality-Control Group

EXPERIMENTAL

Radiologists enrolled in this arm receive access to the study AI radiology report quality-control system within the routine reporting workflow. After report submission, the system performs automated pre-review and may present alerts and suggested revisions. Radiologists retain final decision authority.

Device: Radiology Report GPT (AI-Based Radiology Report Quality-Control System)

Usual Workflow Control Group

NO INTERVENTION

Radiologists enrolled in this arm continue the existing RIS reporting and reviewing workflow without access to the study AI quality-control alerts or suggested revisions.

Interventions

Radiology Report GPT (AI-Based Radiology Report Quality-Control System) DEVICE

Radiology Report GPT is an AI-based pre-review and quality-control system for radiology reports. After a report is submitted, the system may identify potential textual/terminology errors, patient-sex inconsistencies, formatting or numerical errors, examination-type inconsistencies, findings-impression inconsistencies, lesion location/count discrepancies, omissions between findings and impression, and missing or incorrect diagnostic classifications. When a potential issue is detected, the system presents an alert and suggested revision. The radiologist remains responsible for the final report and may accept, modify, or ignore the suggestion. The system does not automatically edit or sign the report. The exact software/model version and the trial deployment mode will be locked and documented before intervention activation.

AI Quality-Control Group

Eligibility Criteria

• Age: 18 Years+
• Sex: all
• Healthy Volunteers: Yes
• Age Groups: Adult (18-64), Older Adult (65+)

You may qualify if:

• Radiologist working in the Department of Radiology at Tongji Hospital.
• Participated in routine radiology reporting during the pre-intervention reference period.
• Reporting work includes at least one target imaging examination type/workflow covered by the study AI quality-control system (CT, MR, and/or CR as applicable).
• Has a valid physician identifier that permits linkage of radiology reports, workflow timestamps, and randomized study assignment.
• Able and willing to provide informed consent.

Sponsors & Collaborators

• Huazhong University of Science and Technology: lead

Study Design

• Study Type: interventional
• Phase: not applicable
• Allocation: RANDOMIZED
• Masking: SINGLE
• Who Masked: OUTCOMES ASSESSOR
• Purpose: HEALTH SERVICES RESEARCH
• Intervention Model: PARALLEL
• Sponsor Type: OTHER
• Responsible Party: PRINCIPAL INVESTIGATOR
• PI Title: Associate Professor

Model Details: Individual radiologists are randomized to AI quality-control access or the usual reporting workflow. The unit of randomization is the radiologist. For reports that involve a reporting and a reviewing radiologist, report-level exposure is classified by the randomized assignments of both radiologists, producing a 2×2 reporter-AI-by-reviewer-AI structure.

Study Record Dates

• First Submitted: October 3, 2026
• First Posted: October 8, 2026
• Study Start: October 1, 2026
• Primary Completion (Estimated): November 1, 2026
• Study Completion (Estimated): January 1, 2027
• Last Updated: October 13, 2026

Record last verified: 2026-10