返回Google圖書搜尋

Using Intraoperative Audio to Predict the Presence of Surgical Errors in Operating Room Black Box Recordings

出版	University of Toronto, 2016
URL	http://books.google.com.hk/books?id=SFVRzwEACAAJ&hl=&source=gbs_api

註釋Recording and review of surgeries can increase safety by allowing researchers to identify errors and analyze their root causes. Watching a surgical can be time consuming. There is a need to make this process more efficient. The research aims to use intraoperative audio to create a machine learning algorithm that can identify times in a surgical recording where an error has been committed. Audio can create effective features due to subtleties in communication, known to change under times of stress. An alternating decision tree is generated using 19 features and a training set of 31 surgeries. The algorithm was tested on 21 surgeries, yielding a sensitivity of 80.8% and overall accuracy of 41.1%. By changing algorithm parameters, an accuracy of 68% was also achieved. The algorithm demonstrates the feasibility of a machine learning algorithm for surgical error prediction. The algorithm should be tested with raters to test its helpfulness.