ThyroCheck AI walks laboratory values through a documented preprocessing pipeline and a compact 1D convolutional neural network, trained on the official UCI Thyroid Disease dataset, to produce an instant, conservative screening pattern - never a diagnosis.
Interpretable inputs drawn directly from the trained model's feature set.
Fill the manual form or upload a CSV batch of fictional screening records.
Saved medians, encoders and scalers (fit on training data only) transform your input.
The cached model produces class probabilities in milliseconds - no retraining, no waiting.
You get a screening-pattern likelihood, confidence, completeness and top contributors.
Each screening is turned into an ordered vector of 49 processed features (scaled lab values, encoded clinical flags, and measured/not-measured indicators). A small 1D convolutional network slides a short filter across this fixed ordering to learn combinations of nearby features, then pools and classifies.
Important: this is a 1D-CNN over a processed feature vector, not a time-series model - thyroid lab attributes are not treated as a sequence over time.
Academic dataset performance - not clinical validation.
Circular probability indicator on every result
Every result shows how complete your submitted data was.
Create a free account and run your first instant screening in under a minute.
Start Screening