Artificial Intelligence in Neurosurgery Anesthesiology: Clinical Applications and Ethical Limitations
Objective:
To explore the possibility of artificial intelligence in neurosurgery
Background:
This perspective analysis examines the recent use of artificial intelligence in anesthesiology and resulting limitations. We show that AI in neurosurgery is used to uncover the interrelationship between drug dose and patient interaction and how pivotal a role it plays in chart review for predicting adverse events & fatal interactions, underlining the need of this research on an urgent basis.
Design/Methods:
We searched medical databases for randomized controlled trials and non-randomized controlled trials that discuss the use of artificial intelligence in anesthesiology. Studies of at least 12 weeks duration were included. We will use Covidence and SciWheel to select and record data that meets the eligibility criteria. At least two researchers review any data that comes into question regarding validity or bias, note it, and address it accordingly.
Results:
A review of 25 studies identified several dominant themes among AI-driven anesthesiology innovations, such as drug dosing algorithms and adverse event prediction in the operating room (accurate at rates up to ~85%). Nonetheless, constraints do persist with incomplete biomechanical modeling leading to inaccurate nerve tracking (30% of studies), suboptimal needle localization (error rate 12-15%), and human manual data inputs display issues. Worries that biases related to patient.
Conclusions:
The use of AI to identify patterns between drug dose and patient interaction is a development, yet its review roles in charts for early prediction of event/fatal interactions only emphasizes the urgent need this research field clearly require. While AI has the opportunity to improve neurosurgery significantly, one must recognize its shortcomings at present being limited to providing global anesthesia care for all patients, inaccuracies in nerve tracking and needle localization technologies as well as difficulties with manual data entry required for surgical outcome predictions; potential reinforcement of implicit bias based on patient demographics.
Disclaimer: Abstracts were not reviewed by Neurology® and do not reflect the views of Neurology® editors or staff.