Transcranial Doppler and Artificial Intelligence Prediction of Postoperative Delirium in Heart Valve Surgery

Postoperative delirium is a frequent complication after cardiac surgery. This prospective observational study will evaluate whether cerebral blood flow velocity measured by transcranial Doppler before and during elective heart valve surgery is associated with delirium during the first five postoperative days. Bispectral index, near-infrared spectroscopy, and arterial blood gas data will also be recorded. Delirium will be assessed daily by an independent evaluator who is blinded to the intraoperative measurements and artificial intelligence predictions, using CAM-ICU for mechanically ventilated patients or the Turkish 3D-CAM for patients on the ward. The study will also compare zero-shot delirium risk predictions from five general-purpose large language models (Claude, GPT, Gemini, Grok, and Kimi) with the blinded clinical assessments. The models will not be trained on data from the study cohort.

Inclusion Criteria: * Age 18 years or older * Scheduled for elective heart valve surgery * Undergoing surgery under general anesthesia according to the study's standardized anesthesia and analgesia protocol * Adequate signal quality for intraoperative BIS, NIRS, and transcranial Doppler measurements * Written informed consent provided by the participant or legally authorized representative Exclusion Criteria: * Emergency or salvage surgery * Redo or reoperation cardiac surgery * Preoperative Standardized Mini-Mental State Examination score below 24 of 30, known dementia, active delirium, or psychotic disorder * Stroke or transient ischemic attack within the previous 3 months * Postoperative prolonged intubation or requirement for extracorporeal membrane oxygenation or an intra-aortic balloon pump * Parkinson disease, epilepsy, multiple sclerosis, or another neurodegenerative disorder * Carotid stenosis detected on preoperative carotid Doppler ultrasonography