An observational study in Thoracic Diseases, Chest CT Scan and Artificial Intelligence (AI) in Diagnosis, sponsored by Shanghai Zhongshan Hospital. Not yet recruiting at 2 sites in China. Open to participants aged 18 Years and older. Per ClinicalTrials.gov, last updated 2026-06-11.
Sponsored by Shanghai Zhongshan Hospital · Observational
The goal of this observational study is to learn if an AI assistant tool can help doctors who read chest CT scans (called radiologists) write their reports faster and just as well or better. Chest CT scans are common pictures taken of the inside of the chest to help with diagnosis. The main questions the study aims to answer are: (1) Does using the AI tool save radiologists time when writing their reports? (2) Are the final reports written with the AI tool's help as good as or better than reports written without it? To answer these questions, researchers will compare two time periods at several hospitals. They will look at how long it took to write reports and how good the reports were, both from a time before the AI tool was available and from a time after it was in regular use. In this study, radiologists will use the AI tool as part of their normal daily work. The tool is built into the computer system they already use to look at scans. Researchers will then measure the time and quality of the reports produced during their regular shifts.
Here we provide a summary of the study's methodological framework, including a description of the AI system under evaluation, key quality control measures, and the data analysis plan. Comprehensive details regarding the full protocol, including eligibility criteria and outcome measures, can be found in the other modules of the study protocol.
72 studies on the registry are indexed under Thoracic Diseases; 13 are open to participants now.
This study's planned enrollment of 75 is below the median of 122 across 19 observational studies indexed under Thoracic Diseases.
Browse Thoracic Diseases studies →Shanghai Zhongshan Hospital is the lead sponsor of 652 studies on the registry; 291 are open to participants now.
Counted across the registry records on this site, refreshed daily.
The study population is defined as follows:
The study participants include both the radiologists whose performance is evaluated and the chest CT scans they interpret. Eligibility criteria are defined for both.
1. Inclusion Criteria
1.1 For Radiologists
1.2 For Chest CT Scans
2. Exclusion Criteria
2.1 For Radiologists:
2.2 For Chest CT Scans
Underlined terms in the criteria open these too. Explained by clinicaltrials.gg, not part of the registry record. The study team decides who can take part.
A biopsy takes a small piece of tissue to examine. “Archival tissue” is a sample already taken, for example at diagnosis, and kept by the hospital.
Whether a stored sample will do depends on when it was taken, how much there is and what the study needs. Ask whether the study accepts your stored sample or needs a new biopsy.
Biopsy and archival tissue
A biopsy takes a small piece of tissue to examine. “Archival tissue” is a sample already taken, for example at diagnosis, and kept by the hospital.
Whether a stored sample will do depends on when it was taken, how much there is and what the study needs. Ask whether the study accepts your stored sample or needs a new biopsy.
Explained by clinicaltrials.gg, not part of the registry record. The study team decides who can take part.
This study employs a single-arm, within-subjects design. A cohort of radiologists will be followed through two sequential practice phases: (1) Baseline (Control) Phase: Participants interpret and report on chest CT scans using their standard clinical workflow without AI assistance. (2) AI-available phase: The same participants interpret and report on a different set of chest CT scans with the integrated AI-assisted reporting system activated in their workflow.
Device: An AI-assisted reporting system integrated into the clinical workflow, providing automated draft generation to assist with chest CT interpretation
The intervention under evaluation is an AI-assisted diagnostic reporting system, integrated directly into the radiologists' workflow. The system analyzes the CT images in real time using an AI model and automatically generates a structured, preliminary radiology report draft. The interpreting radiologist reviews this AI-generated draft, which is presented within their familiar reporting interface. The radiologist then actively edits, confirms, supplements, or overrides the draft content as necessary before finalizing and signing the report. This intervention is distinguished from other AI tools by its focus on end-to-end reporting efficiency via integrated draft generation within the radiologist's classic workflow. It moves beyond simple abnormality detection or highlighting by generating a complete, structured narrative report draft, aiming to reduce dictation/typing time and minimize oversight of findings.
Change in Average Image Interpretation Time
Comparison of the average time taken by participating radiologists to complete standard chest CT interpretation tasks, measured both with and without use of the automated interpretation tool. The time will be recorded from the start to the completion of each individual reading case.
Time frame: Time of interpretation will be collected once the data become fully available (generally within 2 weeks after the planned primary completion date). Final aggregated analysis will be completed within 3 months after the collection of potential confounders.
Change in Chest CT Report Quality Score
Comparison of the subjective quality of chest CT reports written with and without automated tool support. Blinded external experts will evaluate the subjective quality of all sampled reports using a 10-point rating scale, with scores ranging from 1 (poorest quality) to 10 (highest quality).
Time frame: Reports will be distributed to external experts for scoring once the data become available, with scoring results returned within 7 days. Final aggregated analysis will be completed within 3 months after the collection of potential confounders.
Radiologist Editing Intensity on AI-Generated Report Drafts
This measure quantifies the extent of modification a radiologist makes to the initial draft report generated by the AI system. Editing intensity will be algorithmically calculated for each report in the with-AI period. A common method is the normalized edit distance (e.g., Levenshtein distance) or the percentage of text modified between the AI-generated draft and the radiologist's final signed report.
Time frame: Report texts will be collected once the data become fully available (generally within 2 weeks after the planned primary completion date). Final aggregated analysis will be completed within 3 months after the collection of potential confounders.
Plan to share: No — This is an observational study analyzing aggregated, de-identified operational metrics (e.g., radiologist efficiency, report quality scores) derived from routine clinical practice. The data are not collected as part of a prospective clinical trial and are not structured for independent analysis. Findings will be disseminated through peer-reviewed publications.
Start is estimated for Jun 2026; primary completion for Aug 2026, 2 months later.
How these figures are worked out, and which studies are left out of them.
This study is not yet recruiting, as verified in Jun 2026. You cannot join it, but the record below documents what was studied.
Get an email when the registry record changes — status, dates, results — or when someone posts here.
Sign in to followQuestions and observations about this study, from anyone following it. Not medical advice, and not a channel to the study team — their contact details are on the registry record.
Sign in to join the discussion. Reading takes no account; posting does. You choose a display name, and a pseudonym is the default.
Nothing here yet. If you are running this trial, taking part in it, or weighing whether to, this is the place to say so.