Feline pilot
Now enrolling
Model
v1.6
A feline USG screening pilot, for GPs who will log the outcome
Five bloodwork values in. A urine specific gravity prediction out, with the reasoning shown. It is a screening prompt to decide whether a cat needs a urinalysis — never a replacement for running one. Prediction runs on cloud inference, so it needs a connection.
One anonymized
case
A wellness panel, a flag, an early catch
One anonymized case from the originating feline-only practice
A senior cat came in for a routine senior-wellness panel. Bloodwork was unremarkable apart from a mildly elevated BUN; creatinine was within reference. These five inputs went into the USG Predictor (v1.6).
Inputs — 5
- BUN
- 42 mg/dL
- Creatinine
- 1.5 mg/dL
- Hemoglobin
- 12.4 g/dL
- Abs. lymphocytes
- 2,700 /µL
- Patient age
- 13 years
Model output — adequacy score, 0 to 1
0.36
Higher means more likely to be adequately concentrated, so a score this low put the cat on the flagged side. The report calls the score uncalibrated — a ranking signal, not a probability. A prompt to collect urine, not a diagnosis.
- USG, measured
- 1.026
- Diagnosis, by the veterinarian
- CKD Stage 1 (IRIS)
On the flag, the veterinarian collected urine with owner consent. USG came back at 1.026 — inadequate concentration. The cat started an early kidney support diet before becoming symptomatic. The math gave the vet a second look. The vet made the call.
What you would
be validating
The numbers as they stand, including the unflattering one
Model v1.6, April 2026, measured on a temporal holdout of 248 cases never used in training. “Adequate” means a urine specific gravity at or above 1.035; below that is impaired. The model report publishes no confidence intervals and no numeric operating threshold, so neither is stated here. Asking you to generate validation data without showing you what is being validated would be the wrong way round. Workup is not FDA- or USDA-cleared as a medical device, and these are a pilot model’s reported figures, not a cleared diagnostic’s.
92%
Sensitivity
Of the 107 cats in the holdout set whose urine turned out to be inadequately concentrated, the model flagged 98.
60%
Specificity
Of the 141 whose urine was adequately concentrated, 84 were correctly left unflagged — so about 4 in every 10 healthy cats get flagged for a urinalysis they did not need. The report calls this a deliberate trade for keeping sensitivity above 90%, and it is the number this pilot exists to improve.
0.86
AUC-ROC
Overall discrimination between the two groups across all thresholds, on the same holdout set.
Methodology, holdout metrics and validation results are public in the v1.6 model report. The report labels the tool for investigational and research use, states it is not validated for clinical deployment, and records that it has been trained and tested on a single practice’s feline cases with no external validation and no prospective outcome data. That is not a caveat we are burying — it is the reason this pilot exists, and what your cases would change.
Protocol
Three steps,
in order
What you would actually be doing
Step 1
Run the predictor
Use the USG Predictor in the clinic portal on feline patients you have the inputs for.
Step 2
Run the USG
Run the urine specific gravity on cats the algorithm flags — or on the ones you would have run anyway. Both are useful data.
Step 3
Log the outcome
In the portal, walk through a short form: UA result → USG value → diagnosis → “would you have caught this without the flag?” 30 seconds per case.
What you get
What you get, per submitting veterinarian
3 months
Each veterinarian who submits at least 5 pilot outcomes earns three months of full app access at no charge, up to your clinic’s seat count.
The clock
Starts at launch on July 13, 2026, not at enrollment, so signing up early does not shorten the reward. After three months the subscription continues at standard pricing unless you cancel.
Boundaries
Read before
you consent
What this pilot is, and isn’t
Workup is not FDA- or USDA-cleared as a medical device. The USG Predictor is a pilot and a screening prompt, not a diagnostic and not a replacement for urinalysis.
The algorithm flags and predicts. The veterinarian diagnoses.
All patient data is anonymized in any public write-up unless your clinic opts in to be named.
No competitor comparisons in our follow-up content.
Your patient data is partitioned by clinic code. Other pilot clinics cannot read it.
Portal access during the pilot is free. The reward — three months of full app access at no charge — begins at launch on July 13, 2026.
Data submitted through this pilot is used for external validation of the existing USG Predictor model and for any resulting peer-reviewed publication. Submitted data is not incorporated into model training datasets.
By submitting data through this pilot, your clinic represents that it has the right to submit this data and that doing so does not violate any third-party agreement.
Apply
Reply within
2 business days
Tell us about your clinic
We’ll reply within 2 business days. After fit confirmation you’ll receive a portal URL, a clinic code, and a short orientation email.
About
RadAnalyzer’s USG Predictor (v1.6) was developed and validated in a feline-only practice. Tabitha Baibos-Reyes, DVM, our clinical lead, uses it in daily feline practice. This pilot extends it to outside clinics for structured outcome data.
Questions outside the pilot scope? Contact us.