Perioperative medicine has emerged as an exceptionally data-rich domain within healthcare. High-resolution physiological data, point-of-care imaging such as ultrasound, and pre-operative CT/MRI scans, alongside real-time video, laboratory results, and free-text clinical notes, generate a continuous flow of information for surgical patients. In tandem, technological advances in deep learning, transformers, and large language models are offering unprecedented capabilities to integrate heterogeneous data streams. These models allow for the prediction of adverse events, potentially hours or days in advance, and enable tailored interventions based on individual physiological states or genetic profiles. Examples like closed-loop vasopressor control from arterial waveforms and comprehensive EHR-derived models for predicting acute kidney injury highlight the promise of AI. Still, significant challenges persist, such as issues with generalizability, bias, regulatory hurdles, and clinical adoption.

This Research Topic aims to consolidate pioneering research that propels AI methodologies into actionable precision perioperative care. Our objectives include showcasing new algorithms and clinical strategies that utilize physiological signals, imaging, and clinical records for individualized patient predictions and therapies. Additionally, we seek contributions that promote open-source tools, curated datasets, and transparency to enhance reproducibility. A platform will be provided to discuss implementation science, ethics, fairness, and regulatory aspects critical for real-world application, while fostering collaboration among diverse stakeholders including anesthesiologists, surgeons, intensivists, radiologists, data scientists, and ethicists.

To gather further insights in perioperative AI applications, we welcome articles addressing, but not limited to, the following themes:

• Intelligent analysis of physiological signals, including deep and self-supervised learning applications

• AI advancements in imaging for perioperative care and real-time surgical assistance

• Development of multimodal EHR and omics models for advanced patient care

• Applications of large language models in perioperative settings

• Innovations in biomarker discovery and anesthetic drug development

• Ensuring fairness, interpretability, and robust validation in AI healthcare applications

Submissions should adhere to TRIPOD-AI, SPIRIT-AI, or DECIDE-AI guidelines where applicable. Authors are encouraged to share code and de-identified data to support research transparency.

Article types accepted include Original Research, Brief Reports, Methods, Review/Mini-Review, Systematic Review, Policy & Practice Review, Technology & Code, Clinical Trial, and Perspective/Opinion manuscripts.

Are you interested in contributing a manuscript to this Research Topic?

Artificial Intelligence for Precision Perioperative Care: Multimodal Predictive Modelling from Physiology, Imaging and Clinical Records