Perioperative Artificial Intelligence Driven Integrated Modeling of Surgeries using Anesthetic, Physical and...

Area of Science:

• Neurosurgery
• Medical Informatics
• Geriatric Medicine

Background:

• Preoperative cognitive status is crucial for surgical outcomes but remains under-researched.
• Clock drawing tests offer a practical method for assessing cognitive function.
• Deep learning can extract detailed features from clock drawings.

Purpose of the Study:

• To evaluate the predictive power of preoperative cognitive status on surgical outcomes.
• To determine if cognitive assessment surpasses traditional risk factors in predicting outcomes.
• To identify specific cognitive and intraoperative features associated with adverse surgical events.

Main Methods:

• Utilized clock drawing tests with a deep learning algorithm to extract 10 constructional features.
• Trained machine learning models with 5-fold cross-validation to predict postoperative outcomes.
• Employed Shapley Additive Explanations for feature interpretability.

Main Results:

• The perioperative cognitive dataset was the best predictor for 12 of 18 surgery-outcome pairs.
• Surgery duration and anesthetic concentration were significant predictors of adverse outcomes.
• Low average blood pressure, high blood pressure variability, and clock size predicted adverse outcomes.

Conclusions:

• Preoperative cognitive status, particularly assessed through clock drawing features like size, is a valuable predictor of surgical outcomes.
• Integrating cognitive assessment into perioperative modeling can enhance risk prediction and healthcare strategies.
• Intraoperative variables like surgery duration and blood pressure management remain critical for patient safety.