Loading prediction engine

Connecting to prediction server...

0s, cold boot

About the craniopharyngioma outcomes model

This tool estimates two outcomes after craniopharyngioma resection: the probability of permanent diabetes insipidus (DI) and the probability of tumor recurrence. Each prediction comes with a SHAP explanation of which patient and tumor features pushed the risk up or down.

A surgical trade-off view sweeps the extent of resection (subtotal, near-total, gross-total) and shows how DI and recurrence risk move against each other, so the cost of a more aggressive resection is visible next to its benefit.

Predicts: Permanent diabetes insipidus and tumor recurrence, each as a probability with a higher or lower risk label.
Inputs: Age, BMI, tumor dimensions (cm), pre-op visual impairment, hydrocephalus, intracranial pressure symptoms, headache, endocrine changes, panhypopituitarism, intraventricular extension, surgical approach (craniotomy or endoscopic), extent of resection, BRAF status and prior radiation.
Outputs: Risk probabilities, a per-patient SHAP dependence plot, and an extent-of-resection trade-off frontier.
Models and validation: DI: blended ridge (calibrated) and CatBoost model, mean AUC 0.770 across 3 folds. Recurrence: isotonic-calibrated naive Bayes, mean AUC 0.787.

Key terms

Research demonstration only. Not for clinical decision-making.