A ready-made cohort of declined lungs — with the CT to second-guess the call.
Every discarded lung that also has a donor chest CT is a labeled test case: the organ was refused, and the image that could confirm or overturn that judgment already exists. This is the validation set at the heart of VitalMatch.
Source: OPTN/UNOS STAR file, Dec 2025 release · lungs recovered-for-transplant then discarded, intersected with the DICOM→STAR donor crosswalk.
1,774
1,774 discarded lungs from 1,029 donors carry a chest CT — about 19% of all discarded lungs. Roughly 48% were declined for image-assessable reasons, making them directly adjudicable.
01 · The study set
From every discard to the ones we can study
9,197
Discarded lungs — full registry
1,774
…with a donor CT (19% of discards)
1,029
Donors in the CT-linked set
~555
Image-assessable declines (48% of coded)
Each of these lungs pairs a donor CT with a hard outcome — declined for transplant — so the model can be tested on the exact decisions it aims to inform. Note this is a floor: only ~9,300 donors are currently crosswalked to CT, so more discarded lungs likely have images not yet linked.
02 · Why they were declined
Nearly half turn on organ quality a CT can see
Discard reasons for the CT-linked lungs (coded n = 1150). Poor organ function, diseased organ, and anatomical abnormality — all image-assessable — are highlighted.
Discard reason — CT-linked lungs (count)
■ image-assessable · "Other, specify" muted (no coded reason).
03 · Could they have been used?
The discarded lungs weren't obviously worse
On donor oxygenation — the primary objective acceptability criterion — discarded and transplanted lungs barely separate: median P/F 400 vs 444. The distributions overlap heavily, so gas exchange alone doesn't explain the discards.
Donor P/F ratio distribution (% of donors) Discarded Transplanted
~1,461
declined for “poor function” despite P/F ≥ 300
~456
non-quality declines (no recipient / unsuitable in OR)
5,434
discards meeting P/F ≥ 300 — loose upper bound
How many could have been used? An estimate, not a count. Between the documented contradictions (~1,461 lungs declined for poor function despite acceptable gas exchange) and the loose upper bound (5,434 met P/F ≥ 300), a large fraction of discards is objectively questionable. But P/F is necessary, not sufficient — a good ratio doesn't rule out infiltrate, edema, or secretions. That is exactly why adjudication needs the image, not just the number — and why the CT-linked subset above is the set where the question is actually answerable.
04 · The hard cases
These are the marginal donors
The CT-linked discard cohort skews toward exactly the borderline organs where the accept/decline call is hardest — and where an objective read has the most to add.
41% are DCD donors (vs ~7% of all lung donors), and donor age spreads evenly across young, middle, and older tiers. This is not a set of obvious discards — it is the judgment-call population, imaged and labeled, waiting to be adjudicated.