Anesthesia Information Management Systems (AIMS) facilitate the collection of patient data during the perioperative period. These tools interface with intraoperative patient monitors, and some can read and write data to and from hospitals’ clinical repositories (Ehrenfeld & Rehman, 2010). As more advanced artificial intelligence (AI) technologies are developed, they are being used in conjunction with AIMS to solve various problems in anesthesia and health care.
In general, hospitals benefit from AIMS even without the assistance of AI. Since anesthesia “fundamentally relies on the timely collection […] of accurate information,” electronic systems’ performance in “both the capture and interpretation” of increasingly complex data makes AIMS naturally effective tools (Ehrenfeld & Rehman, 2010). Of course, AIMS help provide an accurate after-the-fact record of patients’ data during an operation. In addition, they have also been shown to improve patient safety by making it easier for medical professionals to provide proper care during operations. Some AIMS even provide automatic alerts for incoming clinical events, such as drug interactions or allergic reactions.
AI implementations with Anesthesia Information Management Systems present a logical next step beyond simple data collection and analysis. Compared to traditional computational methods, AI can better interpret complex or subtle shifts within data sets. One open problem in anesthesia, which AI can help address, is creating a measure for the depth of anesthesia (DOA).
It can be very difficult to understand a patient’s DOA, since the interactions between anesthetic drugs and patients’ nervous systems are incredibly complex. However, an accurate measure of DOA would likely become an important tool in the administration of proper anesthetic care (Liu et al., 2015). It would not only help ensure patients were properly anesthetized but would also help guarantee that patients would recover quickly and safely from the effects of anesthetics once medical procedures had concluded.
One relatively successful AI implementation to measure DOA relies on an artificial neural network (ANN), a group of connected nodes which communicate to perform complicated tasks. In this metric, the ANN helps account for even more complex aspects of AIMS-collected EEG data than previous metrics could. The new algorithm is designed to better respond to the change in data over “different [time] scales” (Liu et al., 2015). This so-called “multi-scale entropy” (MSE) measure dramatically outperforms the widely used DOA metric SampEn, which does not account for multiple time scales. The authors conclude that their new approach could be “very useful for accurate and robust measurement of DOA” (Liu et al., 2015).
Furthermore, several studies indicate that AI could be helpful in predicting adverse events during operations. One group of researchers developed a machine learning (ML) model “that is able to classify [hypoxia during anesthesia] on a level that resembles the mutual agreement between human experts” (Sippl et al., 2017). The authors indicate that the techniques used to develop their model could “be applicable to other types of time series [data],” and that future research on these methods might involve “more advanced preprocessing approaches […] such as automated feature learning” (Sippl et al., 2017). This kind of preprocessing would save programmers work in building ML models to rapidly interpret AIMS data and predict even more varieties of adverse events.
In conjunction with AI, AIMS provide a powerful tool with which medical professionals can provide surgical and intensive care. The combination of the two technologies can improve patient safety during anesthesia. Furthermore, AI can help mitigate the risks associated with adverse intraoperative events by interpreting AIMS-collected patient data and performing real-time analysis. Due to the effectiveness and accuracy of AIMS in collecting patient data, the implementation of AI has resulted in these and other “concrete and promising findings […] in the domains of intraoperative medicine” (Cote & Kim, 2019). As AI – a highly-studied frontier in the field of computer science – progresses, researchers in the medical field will likely leverage new computational techniques to ensure that the standards of intraoperative care continue to rise.
References
Cote, C. D., & Kim, P. J. (2019). Artificial intelligence in anesthesiology: Moving into the future. University of Toronto Medical Journal, 96(1), 33 – 36–33 – 36. http://utmj.org/index.php/UTMJ/article/view/1150
Ehrenfeld, J. M., & Rehman, M. A. (2010). Anesthesia information management systems: a review of functionality and installation considerations. Journal of Clinical Monitoring and Computing, 25(1), 71–79. https://doi.org/10.1007/s10877-010-9256-y
Liu, Q., Chen, Y.-F., Fan, S.-Z., Abbod, M. F., & Shieh, J.-S. (2015). EEG Signals Analysis Using Multiscale Entropy for Depth of Anesthesia Monitoring during Surgery through Artificial Neural Networks. Computational and Mathematical Methods in Medicine, 2015, 1–16. https://doi.org/10.1155/2015/232381
Sippl, P., Ganslandt, T., Prokosch, H.-U., Muenster, T., & Toddenroth, D. (2017). Machine Learning Models of Post-Intubation Hypoxia During General Anesthesia. https://doi.org/10.3233/978-1-61499-808-2-212