<p>Adjustable pressure cerebrospinal fluid (CSF) shunt valves are widely used in the treatment of hydrocephalus. CSF shunt dysfunctions can manifest with diverse symptoms, often requiring further diagnostic evaluation. Head computed tomography (CT) is frequently used as an initial diagnostic tool. Accurate identification of the current shunt valve setting is crucial for patient management; however, interpretation on CT is difficult due to three-dimensional imaging, metal artefacts, and limited spatial resolution. We therefore developed an artificial intelligence (AI)-based model to automatically assess shunt valve settings in CT scans. We collected 391 head CT scans from patients with CSF shunts featuring a Certas Plus valve. Shunt settings were extracted from medical records and verified on imaging. A 3D U-Net was trained to segment radiopaque valve components, from which the valve setting was inferred. The model successfully segmented valve components in 97.3% of test cases and correctly predicted the exact or an adjacent setting in 96% of cases. The segmentations enable clinicians to interpret and verify the prediction. Our study demonstrates the feasibility of AI-based detection of programmable shunt valve settings in CT scans. The proposed model reliably identifies Certas Plus valve settings and holds promise as a clinical support tool in shunt diagnostics.</p>