Abdominal CT · 18 structures · 146 findings
Expert-level generalist AI for abdominal CT
Upload a contrast-enhanced abdominal CT volume or DICOM series and get per-finding scores in about a minute. Free for non-commercial research use (CC BY-NC-SA 4.0).
Score a case1. Upload
One volume file (.nii.gz, .nrrd, .mha) or a full DICOM series — drag, drop, done.
2. Automatic analysis
We prepare your scan, then our AI reads it. You get results in about 90 seconds.
3. Review findings
Top-10 bars plus a sortable table of every scored finding (up to 146).
Questions & answers
What does RADAR do?
RADAR scores contrast-enhanced abdominal CT scans across 18 structures and up to 146 findings. You get per-finding scores — top-10 bars plus a searchable, sortable table — not generated report text or a confirmed diagnosis.
What files can I upload?
Either one volume file (.nii, .nii.gz, .nrrd, .mha) or one full DICOM series (multiple .dcm or extensionless files, or a .zip — nested folders are OK; the series with the most slices is scored). A single .dcm is one slice, not a volume. Header/raw pairs (.mhd, .nhdr, .hdr, .img, .raw) must be converted to a supported volume first. A DICOM series needs at least 8 slices.
How big can files be, and how long does scoring take?
Browser limit is 500 MB per file; on the public demo the free edge refuses bodies over about 100 MB, so the site pre-checks remote totals at about 95 MB. Scoring usually takes about a minute (around 90 seconds), one case at a time — if the queue is full you get a queue-full error and should retry.
Is this a medical diagnosis?
No. Research assistance only (CC BY-NC-SA 4.0). Scores must not substitute for clinical review.
Is my scan private? Do you keep my data?
Only upload scans you are allowed to share — ideally fully anonymized, with no patient name, ID, or dates. Your files are used only to compute scores and are deleted afterwards: not kept, reused, or shared. The public demo has no accounts and is open to everyone, so never submit anything identifiable or protected.
Can I trust the scores?
Treat them as a second opinion for research, not a diagnosis. This is a free demo of the published RADAR model (see the model link below) for non-commercial research — it only reads contrast-enhanced abdominal CT, it can be wrong, and it never replaces an expert's review. If a case fails, it is usually the file: one full series, anonymized, under the size limit (see above) — fix that and try again.