PneumothoraxNCT07870928
The Impact of Artificial Intelligence-Assisted X-ray Interpretation on Radiologists' Diagnostic Performance
An observational study of Pneumothorax and Intestinal Obstruction, run by The Third Affiliated Hospital of Southern Medical University.
Trial summary
This study tests whether an artificial intelligence (AI) tool helps radiologists read X-rays more accurately. Radiologists read X-rays to find diseases such as pneumonia, bone fractures, and abdominal problems. AI is a computer program trained on many past X-rays. Read more about this study
- Who is it for?
- Ages 18 years and olderHealthy volunteers can take part
- Conditions the study is for
- Pneumothorax, Intestinal Obstruction, Spondylolysis
- How many people
- 500 participants
- Where
- No sites listed yet
- Study dates
- Oct 2026to Jul 2027
- Phase
- Observational
- Are there results?
- Not yet: the study is under way
Key requirements
Who can take part, as listed by the sponsor. Only the study team can confirm whether you are eligible, so take this list to your doctor.
- Ages 18 years and older
- All sexes
- Healthy volunteers accepted
- Adult, Older adult
Study population: This study enrolls 10 radiologists from multiple hospitals as readers. They read retrospective, de-identified X-ray images of the chest, abdomen, and lumbar spine, collected from participating hospital image archives and public datasets. The image cases are from adult patients (age 18 years or older) with common X-ray findings across chest, abdominal, and spinal diseases. No patients are prospectively enrolled; all images are reused from routine clinical care after de-identification.
Who may be able to join
- Participants (radiologists): actively practicing radiologists at participating hospitals, stratified by hospital and professional seniority, trained for the study reading task, and willing to provide written informed consent.
- Cases (X-ray images): adult patients aged 18 years or older who underwent X-ray imaging of the chest, abdomen, or lumbar spine for clinical indications.
- Cases with complete basic clinical information.
- Cases with an established gold-standard diagnosis, defined as the comprehensive clinical diagnosis or a consensus opinion from at least two senior radiologists (deputy chief physician or above).
Who cannot join
- Cases with poor image quality that prevents reliable interpretation.
- Cases without a clear gold-standard diagnosis or with missing key clinical information.
- Duplicate cases.
- Cases in which the patient recently received radiotherapy, chemotherapy, or other treatment that would significantly alter imaging appearance and confound diagnosis.
- Cases judged by the investigators as unsuitable for the study for any other reason.
- Radiologists: no additional exclusion criteria beyond the inclusion requirements.
Trial locations and contacts
No sites are listed yet. Contact the study team or a site near you to ask about taking part.
About this study
Official title: The Impact of Artificial Intelligence-Assisted X-ray Interpretation on Radiologists' Diagnostic Performance: A Multicenter, Multi-Reader, Multi-Case (MRMC) Randomized Cross-Over Study
This study tests whether an artificial intelligence (AI) tool helps radiologists read X-rays more accurately.
Radiologists read X-rays to find diseases such as pneumonia, bone fractures, and abdominal problems. AI is a computer program trained on many past X-rays. It may help radiologists spot diseases faster and more correctly. This study compares reading X-rays with the AI tool to reading X-rays without it.
Read more
Who takes part
- Ten radiologists from several hospitals.
- X-ray images from about 300 to 500 past hospital visits.
- No new X-rays are taken.
- All names and ID numbers are removed from the images.
What happens
- Each radiologist reads the same set of X-rays twice.
- One time, the radiologist reads without the AI tool.
- The other time, the radiologist can see the AI result.
- Half start with no AI and then switch to AI. The other half start with AI and then switch to no AI.
- There is a four-week break between the two times.
- The researchers measure how accurate each radiologist is. They also time how long each image takes.
Possible benefits and risks
- The radiologists may get no direct personal benefit.
- The results may show whether AI can help doctors read X-rays better in the future.
- This is a low-risk study. The AI tool never tells a patient a diagnosis directly.
- A qualified radiologist always makes the final diagnosis.
- Patient data is stored on a locked, encrypted server.
Read the detailed description
This is a multi-reader multi-case (MRMC) crossover diagnostic accuracy study using retrospective, de-identified X-ray images. Ten radiologists read the same set of cases under two conditions - unassisted reading and AI-assisted reading - in random order (AB/BA sequence), with a four-week washout between stages. Each radiologist serves as their own control. No patients are prospectively enrolled, and no new imaging procedures are performed.
- Conditions
- Pneumothorax, Intestinal Obstruction, Spondylolysis, Renal Calculus
- Medical subject headings
- Pneumothorax, Intestinal Obstruction, Spondylolysis, Kidney Calculi
What participants receive
What participants receive. In a randomised study each person is assigned to one group by chance.
| Study group | Role | What they receive |
|---|---|---|
| Unassisted Reading | Radiologists read X-ray images using their usual clinical workflow, without access to any AI tool output. They record their diagnosis, confidence level, and reading time. | |
| AI-Assisted Reading | Radiologists read the same X-ray images while having access to the output of an X-ray imaging AI tool, including its diagnostic suggestions and predicted probabilities. The radiologist reviews the AI output and makes the final diagnosis, then records confidence level and reading time. |
What is being measured
The primary outcome is the main question the study is designed to answer.
Macro-averaged Diagnostic Accuracy of Radiologists' Single-Disease Binary Detection (Balanced Accuracy）
Diagnostic performance of radiologists on single-disease binary detection tasks for four common X-ray diseases. Accuracy is quantified as macro-averaged balanced accuracy (B-ACC) across the four diseases, comparing unaided reading versus AI-assisted reading. Each binary decision is compared against an expert-adjudicated reference standard established from clinical, imaging, and follow-up information.
Measured: Baseline (Stage 1 reading) and Week 4 (Stage 2 reading, after a 4-week washout period)
Secondary outcomes (4)
Per-disease Sensitivity and Specificity
Disease-specific sensitivity and specificity of radiologists' binary diagnostic decisions under unaided versus AI-assisted conditions, estimated using generalized linear mixed-effects models.
Measured: Baseline (Stage 1 reading) and Week 4 (Stage 2 reading, after a 4-week washout period)
Single-Disease Diagnostic Reading Time
Time in seconds from case presentation to submission of the binary diagnostic decision, recorded by an independent timer per case. Compared between unaided and AI-assisted reading conditions.
Measured: Baseline (Stage 1 reading) and Week 4 (Stage 2 reading, after a 4-week washout period)
Inter-Reader Diagnostic Agreement (Kappa)
Inter-reader consistency of binary diagnostic decisions among the 10 participating radiologists, assessed by Kappa statistics, under unaided versus AI-assisted conditions.
Measured: Baseline (Stage 1 reading) and Week 4 (Stage 2 reading, after a 4-week washout period)
Rate of Diagnostic Decision Changes With AI Assistance
Proportion of cases for which each radiologist's binary decision differed between the unaided session and the AI-assisted session, classified as correct-to-incorrect or incorrect-to-correct changes.
Measured: Baseline (Stage 1 reading) and Week 4 (Stage 2 reading, after a 4-week washout period)
Timeline
From the start of the study to its completion; the mark is today. Estimated dates are the sponsor’s plan.
- First posted13 Oct 2026
- Study startOct 2026 (estimated)
- Primary completionJul 2027 (estimated)
- Study completionJul 2027 (estimated)
Resources
- Official record on ClinicalTrials.govThe registry entry this page is built from, with its full change history.
- Other trials for PneumothoraxStudies of the same condition, open and under way.
- What the trial phases meanHow a treatment moves from first tests in people to approval.
More details
For doctors and researchers: how the study is designed, who runs it, and its registry IDs.
Study design
- Study type
- Observational
- Observational model
- Other
- Time perspective
- Retrospective
- Enrolment
- 500 participants (estimated)
Sponsor and investigators
- Lead sponsor
- The Third Affiliated Hospital of Southern Medical University
- Sponsor type
- Other government
- Responsible party
- Principal investigator, Junyan Li, The Third Affiliated Hospital of Southern Medical University
IDs, dates and publications
- ClinicalTrials.gov ID
- NCT07870928
- Sponsor's protocol ID
- 2026-ER-126
- Status last verified
- Oct 2026
- First submitted
- 30 Sep 2026
- Last updated
- 15 Oct 2026
- Results posted
- No
- FDA-regulated drug
- No
- FDA-regulated device
- No
- Individual data sharing
- Yes
Study at a glance
From the registry record, last updated by the sponsor on 15 Oct 2026. Recruitment status is as the sponsor reported it: confirm with the study team.
- Official title
- The Impact of Artificial Intelligence-Assisted X-ray Interpretation on Radiologists' Diagnostic Performance: A Multicenter, Multi-Reader, Multi-Case (MRMC) Randomized Cross-Over Study
- ClinicalTrials.gov ID
- NCT07870928
- Acronym / sponsor ID
- 2026-ER-126
- Conditions
- Pneumothorax, Intestinal Obstruction, Spondylolysis, Renal Calculus
- Study type
- Observational
- Recruitment status
- Not yet recruiting
- Enrolment
- 500 participants (estimated)
- Lead sponsor
- The Third Affiliated Hospital of Southern Medical UniversityOther government
- Study start
- Oct 2026 (estimated)
- Primary completion
- Jul 2027 (estimated)
- Study completion
- Jul 2027 (estimated)
- First posted
- 13 Oct 2026 (actual)
- Results
- Not posted in the source record
Clinical summary
The research question as the sponsor registered it. This describes what the study plans to test, not what it found.
This study tests whether an artificial intelligence (AI) tool helps radiologists read X-rays more accurately.
Radiologists read X-rays to find diseases such as pneumonia, bone fractures, and abdominal problems. AI is a computer program trained on many past X-rays. It may help radiologists spot diseases faster and more correctly. This study compares reading X-rays with the AI tool to reading X-rays without it.
Who takes part
- Ten radiologists from several hospitals.
- X-ray images from about 300 to 500 past hospital visits.
- No new X-rays are taken.
- All names and ID numbers are removed from the images.
What happens
- Each radiologist reads the same set of X-rays twice.
- One time, the radiologist reads without the AI tool.
- The other time, the radiologist can see the AI result.
- Half start with no AI and then switch to AI. The other half start with AI and then switch to no AI.
- There is a four-week break between the two times.
- The researchers measure how accurate each radiologist is. They also time how long each image takes.
Possible benefits and risks
- The radiologists may get no direct personal benefit.
- The results may show whether AI can help doctors read X-rays better in the future.
- This is a low-risk study. The AI tool never tells a patient a diagnosis directly.
- A qualified radiologist always makes the final diagnosis.
- Patient data is stored on a locked, encrypted server.
Detailed description
This is a multi-reader multi-case (MRMC) crossover diagnostic accuracy study using retrospective, de-identified X-ray images. Ten radiologists read the same set of cases under two conditions - unassisted reading and AI-assisted reading - in random order (AB/BA sequence), with a four-week washout between stages. Each radiologist serves as their own control. No patients are prospectively enrolled, and no new imaging procedures are performed.
Study design
- Observational model
- Other
- Time perspective
- Retrospective
- Number of arms
- 2
| Arm | Type | Receives |
|---|---|---|
| Unassisted Reading | – | –Radiologists read X-ray images using their usual clinical workflow, without access to any AI tool output. They record their diagnosis, confidence level, and reading time. |
| AI-Assisted Reading | – | –Radiologists read the same X-ray images while having access to the output of an X-ray imaging AI tool, including its diagnostic suggestions and predicted probabilities. The radiologist reviews the AI output and makes the final diagnosis, then records confidence level and reading time. |
Who may qualify
- Ages
- 18 years and older
- Sex
- All
- Healthy volunteers
- Accepted
- Age groups
- Adult, Older adult
- Study population
- This study enrolls 10 radiologists from multiple hospitals as readers. They read retrospective, de-identified X-ray images of the chest, abdomen, and lumbar spine, collected from participating hospital image archives and public datasets. The image cases are from adult patients (age 18 years or older) with common X-ray findings across chest, abdominal, and spinal diseases. No patients are prospectively enrolled; all images are reused from routine clinical care after de-identification.
Key inclusion criteria (4 listed)
- Participants (radiologists): actively practicing radiologists at participating hospitals, stratified by hospital and professional seniority, trained for the study reading task, and willing to provide written informed consent.
- Cases (X-ray images): adult patients aged 18 years or older who underwent X-ray imaging of the chest, abdomen, or lumbar spine for clinical indications.
- Cases with complete basic clinical information.
- Cases with an established gold-standard diagnosis, defined as the comprehensive clinical diagnosis or a consensus opinion from at least two senior radiologists (deputy chief physician or above).
Key exclusion criteria (6 listed)
- Cases with poor image quality that prevents reliable interpretation.
- Cases without a clear gold-standard diagnosis or with missing key clinical information.
- Duplicate cases.
- Cases in which the patient recently received radiotherapy, chemotherapy, or other treatment that would significantly alter imaging appearance and confound diagnosis.
- Cases judged by the investigators as unsuitable for the study for any other reason.
Eligibility is decided by the study team after screening; this overview does not establish that a patient qualifies. Full inclusion and exclusion criteria.
Primary outcome
Macro-averaged Diagnostic Accuracy of Radiologists' Single-Disease Binary Detection (Balanced Accuracy）
Time frame: Baseline (Stage 1 reading) and Week 4 (Stage 2 reading, after a 4-week washout period)
Diagnostic performance of radiologists on single-disease binary detection tasks for four common X-ray diseases. Accuracy is quantified as macro-averaged balanced accuracy (B-ACC) across the four diseases, comparing unaided reading versus AI-assisted reading. Each binary decision is compared against an expert-adjudicated reference standard established from clinical, imaging, and follow-up information.
4 secondary outcomes
Per-disease Sensitivity and Specificity
Time frame: Baseline (Stage 1 reading) and Week 4 (Stage 2 reading, after a 4-week washout period)
Single-Disease Diagnostic Reading Time
Time frame: Baseline (Stage 1 reading) and Week 4 (Stage 2 reading, after a 4-week washout period)
Inter-Reader Diagnostic Agreement (Kappa)
Time frame: Baseline (Stage 1 reading) and Week 4 (Stage 2 reading, after a 4-week washout period)
Rate of Diagnostic Decision Changes With AI Assistance
Time frame: Baseline (Stage 1 reading) and Week 4 (Stage 2 reading, after a 4-week washout period)
Results and safety
This study is still under way. Results are not available in the source record yet.
Research links
- Official record on ClinicalTrials.govLast updated by the sponsor 15 Oct 2026.
- Individual participant dataSharing planned: Yes
Related trials for Pneumothorax
RecruitingObservational study
Study of the Disease Process of Lymphangioleiomyomatosis
Lung Disease, Pneumothorax
RecruitingObservational study
Genetic Analysis of Birt Hogg-Dube Syndrome and Characterization of Predisposition to Kidney Cancer
Kidney Neoplasms, Kidney Cancer
Not yet recruitingPhase 2/3
Stop Air Leak by Talc or Autologous Blood Patch Therapy
Pneumothorax Spontaneous Secondary
Active, not recruiting
Outpatient Management of Primary Spontaneous Pneumothorax: Pigtail Catheter With Unidirectional Valve vs. Exsufflation, Randomized Prospective Study
Pneumothorax Spontaneous Primary
Not yet recruitingObservational study
Clinical Value of Routine Chest Radiography and Point-of-Care Lung Ultrasound Following Ambulatory Thoracic Sympathectomy
Pneumothorax Iatrogenic Postprocedural
Referral and contact
AzTrials does not enrol participants or take referrals. Contact the study team or a site listed below; they decide who can take part.
Questions about this trial
What is this clinical trial studying?
An observational study of Pneumothorax and Intestinal Obstruction, run by The Third Affiliated Hospital of Southern Medical University.
The full registered description is in the About this study section and the official record.
Who may be eligible to take part?
The sponsor lists 4 inclusion and 6 exclusion criteria. Eligibility is decided by the study team after screening, so read the full criteria and contact a listed site.
Is this clinical trial currently recruiting?
The record says: opening soon, as of the sponsor's last update on 15 Oct 2026. Status can change between updates, so confirm with the study contact before making plans.
What are the possible benefits and risks?
They depend on the study and the treatment, and no study can promise a benefit. The study team explains the known risks during informed consent, before anyone joins; the record itself does not list them.
Are results available for this study?
This study is still under way. Results are not available in the source record yet.
How can a doctor refer a patient or contact the research team?
Contact HUANG WeiHong HUANG (+86-13682232823, 13682232823@163.com), or a site near the patient. AzTrials does not enrol participants or take referrals.
What does the primary outcome measure?
Macro-averaged Diagnostic Accuracy of Radiologists' Single-Disease Binary Detection (Balanced Accuracy） (time frame: Baseline (Stage 1 reading) and Week 4 (Stage 2 reading, after a 4-week washout period)). It is what the study was designed to measure, not a finding.
Where can I read the original study information?
Does a listing mean the treatment is approved or proven to work?
No. A listing means the study is registered. It does not show that the treatment is safe, works, or is approved by a regulator.
How often is this information updated?
AzTrials copies changed records from ClinicalTrials.gov every night and reloads the whole registry every week. The sponsor last updated this record on 15 Oct 2026.