Machine Learning for Predicting Neurosurgical Postoperative Cognitive Decline
Description
Background: Postoperative cognitive decline (POCD) in common in neurosurgical patients and significantly impacts functional independence. Several factors correlate to exacerbated POCD, including age, preoperative brain activity, and anesthesia type. However, little research has explored the use of artificial intelligence (AI) to predict POCD, which may personalize and improve patient treatment. This study aims to develop an AI tool which predicts POCD from neurosurgical patients based on patient and procedure characteristics for improved treatment outcomes. Hypothesis: We hypothesize that an AI tool trained on patient characteristics before and after surgery, along with procedure factors, may predict POCD. Methods: An AI tool will be trained and validated on neurosurgical case data to predict POCD risk and cognitive score (n = 200). Training data will include patient factors (demographics, blood test results, cognitive scores, lesion type) and procedure factors (anesthesia type, neurosurgical intervention). Cognitive score will be assessed using standard MoCA and MMSE tests. The tool will predict 7- and 30-day post-operative MoCA and MMSE score. We will validate the tool using a subset of patients (n = 50) and will assess model accuracy. Implications: AI-driven POCD prediction offers a cost-effective approach to personalized patient care. Integrating these tools into clinical workflows may better equip healthcare service providers to identify high risk patients, assess surgical risk, and adjust management, ultimately improving patient outcomes.
Machine Learning for Predicting Neurosurgical Postoperative Cognitive Decline
Background: Postoperative cognitive decline (POCD) in common in neurosurgical patients and significantly impacts functional independence. Several factors correlate to exacerbated POCD, including age, preoperative brain activity, and anesthesia type. However, little research has explored the use of artificial intelligence (AI) to predict POCD, which may personalize and improve patient treatment. This study aims to develop an AI tool which predicts POCD from neurosurgical patients based on patient and procedure characteristics for improved treatment outcomes. Hypothesis: We hypothesize that an AI tool trained on patient characteristics before and after surgery, along with procedure factors, may predict POCD. Methods: An AI tool will be trained and validated on neurosurgical case data to predict POCD risk and cognitive score (n = 200). Training data will include patient factors (demographics, blood test results, cognitive scores, lesion type) and procedure factors (anesthesia type, neurosurgical intervention). Cognitive score will be assessed using standard MoCA and MMSE tests. The tool will predict 7- and 30-day post-operative MoCA and MMSE score. We will validate the tool using a subset of patients (n = 50) and will assess model accuracy. Implications: AI-driven POCD prediction offers a cost-effective approach to personalized patient care. Integrating these tools into clinical workflows may better equip healthcare service providers to identify high risk patients, assess surgical risk, and adjust management, ultimately improving patient outcomes.
https://scholar.uwindsor.ca/we-spark-conference/2025/postersessions/54