Delirium, a sudden and severe disturbance in mental abilities, is a common and dangerous complication in postoperative patients, particularly the elderly. It increases hospital stays, healthcare costs, and mortality rates. Despite its significance, early prediction and prevention remain a challenge in clinical practice. However, recent advances in artificial intelligence (AI) and electroencephalography (EEG) may be combating that issue, as researchers investigate whether harnessing these tools could allow physicians to detect delirium prior to its onset.

EEG is a non-invasive method of monitoring electrical activity in the brain. In the perioperative setting, EEG can be used to assess the depth of anesthesia. Interestingly, studies have found that specific EEG patterns, such as decreased alpha power, burst suppression, and slow-wave activity, are associated with higher risks of postoperative delirium. These neural signatures often precede the clinical symptoms of delirium, making them valuable predictive markers ^1,2.

By training AI models on labeled EEG data from thousands of patients, researchers are creating algorithms to try to detect subtle brain activity changes that are predictive of future delirium episodes. Deep learning algorithms, for example, can analyze intraoperative EEG recordings to forecast which patients are most likely to experience delirium postoperatively. Some models have achieved prediction accuracies exceeding 80%, outperforming traditional clinical risk scoring methods. These AI systems tend to take into account a variety of EEG features (like spectral entropy, coherence, and power spectral density) to generate individualized risk profiles in real-time ^3–6.

The implications are substantial. Early detection enables targeted interventions, like adjusting anesthetic depth, optimizing postoperative pain control, or initiating early mobilization, to potentially prevent delirium or reduce its severity. For high-risk patients, clinicians can intensify monitoring and proactively manage modifiable risk factors ^7,8.

However, challenges remain in implementing AI models. Not only can the quality and consistency of EEG data across different institutions vary, complicating model generalization, but many existing studies are retrospective, limiting real-world applicability. In addition, there is a need for standardized protocols and validation across diverse patient populations to ensure equity and accuracy ^5,9.

Despite these limitations, the integration of AI with EEG monitoring represents a promising opportunity to improve perioperative care by enabling healthcare providers to detect early signs of delirium. Some hospitals are already piloting AI-driven platforms that integrate with anesthesia monitors to provide real-time risk alerts for delirium. As technology matures and regulatory standards evolve, such tools may become routine in surgical settings ^10–12.

While AI cannot prevent delirium, its ability to detect early EEG changes linked to delirium risk offers a promising frontier in personalized medicine. With continued research, validation, and implementation, AI-powered EEG analysis could transform how clinicians manage cognitive outcomes in surgical patients.

References

1. Williams Roberson, S. et al. Quantitative EEG signatures of delirium and coma in mechanically ventilated ICU patients. Clin Neurophysiol 146, 40–48 (2023). DOI: 10.1088/1741-2552/aac960
2. Wiegand, T. L. T., Rémi, J. & Dimitriadis, K. Electroencephalography in delirium assessment: a scoping review. BMC Neurol 22, 86 (2022). DOI: 10.1186/s12883-022-02557-w
3. Han, C. et al. Machine learning with clinical and intraoperative biosignal data for predicting postoperative delirium after cardiac surgery. iScience 27, 109932 (2024). DOI: 10.1016/j.isci.2024.109932
4. Röhr, V., Blankertz, B., Radtke, F. M., Spies, C. & Koch, S. Machine-learning model predicting postoperative delirium in older patients using intraoperative frontal electroencephalographic signatures. Front Aging Neurosci 14, 911088 (2022). DOI: 10.3389/fnagi.2022.911088
5. Lv, S., Li, J., He, H., Zhao, Q. & Jiang, Y. Artificial intelligence applications in delirium prediction, diagnosis, and management: a systematic review. Artif Intell Rev 58, 269 (2025). DOI: 10.1007/s10462-025-11264-0
6. Jauk, S. et al. Machine learning-based delirium prediction in surgical in-patients: a prospective validation study. Jamia Open 7, ooae091 (2024). DOI: 10.1093/jamiaopen/ooae091
7. Radboud University Medical Center. DElirium prediCtIon in the intenSIve Care Unit: Head to Head comparisON of Two Delirium Prediction Models. https://clinicaltrials.gov/study/NCT02518646 (2017).
8. AI Model Improves Delirium Prediction, Leading to Better Health Outcomes for Hospitalized Patients | Mount Sinai – New York. Mount Sinai Health System https://www.mountsinai.org/about/newsroom/2025/ai-model-improves-delirium-prediction-leading-to-better-health-outcomes-for-hospitalized-patients.
9. Boudewyn, M. A. et al. Managing EEG studies: How to prepare and what to do once data collection has begun. Psychophysiology 60, e14365 (2023). DOI: 10.1111/psyp.14365
10. Morris, M. X., Fiocco, D., Caneva, T., Yiapanis, P. & Orgill, D. P. Current and future applications of artificial intelligence in surgery: implications for clinical practice and research. Front. Surg. 11, (2024). DOI: 10.3389/fsurg.2024.1393898
11. Amin, A. et al. Future of Artificial Intelligence in Surgery: A Narrative Review. Cureus 16, e51631. DOI: 10.7759/cureus.51631
12. Critical Care Innovation: Scientists Improve AI and Rapid-Response EEGs for Delirium Detection | The Healthcare Technology Report. https://thehealthcaretechnologyreport.com/critical-care-innovation-scientists-improve-ai-and-rapid-response-eegs-for-delirium-detection/ (2023).