Completed OBSERVATIONAL NCT07712952

Replaying old radiographs to ask whether it could have been caught earlier — fifteen years of images before an IPF diagnosis, read again by AI — a clinical trial (ClinicalTrials.gov)

Chung-Ang University Hospital Updated 2026-07-20

A retrospective observational study of 175 patients diagnosed with idiopathic pulmonary fibrosis, reaching back through the chest radiographs taken before diagnosis to compare the date AI first detected interstitial opacity with the date a radiologist first reported it. The imaging window runs fifteen years, from January 2010 to April 2025, and the registered start and completion dates fall on the same day.

Trial overview (primary data)

  • StatusCompleted
  • ConditionsIdiopathic Pulmonary Fibrosis, Interstitial Lung Disease
  • InterventionsDEVICE: VUNO Med-Chest X-ray
  • SponsorChung-Ang University Hospital
  • Target enrollment175 participants
  • Period2025-04-30 〜 2025-04-30

Key points

  • A single-center retrospective observational study comparing, in 175 patients with idiopathic pulmonary fibrosis, the date AI detected interstitial opacity against the date a radiologist reported it.
  • The imaging window spans fifteen years, January 2010 to April 2025, reaching back from an April 2025 registry screening.
  • Start and completion dates are both 2025-04-30, so the registered duration is zero days, a signature of retrospective work on data that already exists.
  • The AI decision uses a score cutoff fixed in advance. The comparator is not reading ability but the finding as documented in the record.
  • This describes the design and aims of the study; effectiveness has not been established.

1A question only a backward look can ask

Idiopathic pulmonary fibrosis stiffens the lungs for reasons that stay unknown, and the record states a median survival of only three to five years after diagnosis. Finding it early and starting antifibrotic therapy may change outcomes, yet in practice the diagnosis is often late.

Chest radiography is cheap and available almost everywhere, but its sensitivity for early interstitial opacity is limited, and time can pass without the finding being written down. What this study asks is not whether the AI is accurate. It asks whether, back then, it could have been caught earlier.

2A trial that lasts zero days

Duration of this study0 daysStart date and completion date are both registered as 2025-04-30
Median duration among the 803 records this site holds as of 2026-09-04534 daysBoth dates are recorded for 801 of them
Records that frame the question as catching something earlier30.4% of those same 803

The registered start and completion dates fall on the same day, so the duration comes out as zero. The data already exists. A retrospective study never has to wait for patients, and in exchange it can only ask what the images already taken and the reports already written will support. Set against a median duration of 534 days across the records held here, the speed of this rung and its limits appear together.

3What is compared with what

The comparisonThe date AI first detected interstitial opacityThe date a radiologist first reported it
The criterionA score cutoff fixed in advanceThe wording left in the clinical report
The scopeEvery chest radiograph taken before the diagnosisThe same series of images

The cohort is 175 patients diagnosed with idiopathic pulmonary fibrosis at a single center. Anchored to a registry screening in April 2025, the study reaches back across fifteen years of imaging, from January 2010 to April 2025.

The point is the gap between the day the AI caught the opacity and the day a radiologist wrote it down, with exploratory work on which patient characteristics widen that gap and whether the trajectory of the AI score relates to mortality.

4It is the record that is being evaluated

What stands on the other side of this comparison is not a radiologist's ability to read, but the finding as it was written into the record. Declining to mention a faint opacity that might one day become fibrosis, on a film taken to answer an acute problem, is not an error in the care of that moment. A backward look deals with the difference that appears only when those moments are laid out again as a series. In a prospective trial, that difference can no longer be measured.

Why it matters

A retrospective design can only ask what past images and past records will support, and in exchange it can measure a difference no prospective trial can reach. Laying documentation out as a series to ask what might already have been visible transfers well beyond imaging, to any field where records accumulate. Defining the comparator as what was documented, rather than as ability, is itself a useful move in evaluation design.

FAQ

Why is the duration zero days?
In a retrospective study the images and records used for analysis already exist, so no follow-up period is needed. The registered start and completion dates fall on the same day.
Is this a study about radiologists missing findings?
No. The comparator is the finding as documented in the record, and the design measures how documentation made for the purpose of each visit looks once the visits are laid out as a series.
Does this establish that the AI is useful?
The record sets out the design and aims of the study; it does not report results or establish effectiveness.

Sources (primary)

Source: ClinicalTrials.gov (U.S. NIH/NLM, public domain). This site does not provide medical advice. Verify the latest and exact details with the official source. This site is not endorsed or certified by the NIH/NLM.

#Clinical trials#AI#Healthcare#Respiratory#Imaging
Disclaimer: This site independently summarizes and classifies information based on official data sources. Always verify the latest and accurate information with the official sources. Content on finance, health, legal, and security is information, not advice. This site is not an official website of the U.S. government.