Thyroid Ultrasound Model
Nodule risk stratification, recalibrated.
A research-stage thyroid nodule model trained on Saudi and GCC ultrasound data, where iodine status and malignancy patterns diverge from existing cohorts.
TI-RADS doesn't behave the same everywhere
The most-cited nodule AI was trained on South Korean and Western cohorts. Local iodine status and malignancy mix differ, and thresholds calibrated abroad slip exactly where it matters, on indeterminate nodules.
We're building an annotated ultrasound dataset with Saudi endocrine and head-and-neck programs, training a TI-RADS-aware model, and validating against local cytology outcomes.
- Annotated dataset with partner endocrine units.
- TI-RADS-aware features, locally calibrated.
- Validated against cytology and surgical outcomes.
Who should partner with us
High-volume thyroid imaging services, endocrine surgery programs, and holders of annotated ultrasound data.
Data governance
IRB-governed; data stays in-Kingdom under PDPL.
Partner with us
Research partnerships: IRBs, faculty co-investigators, shared publications.