AI in Anaesthesiology and Critical Care
Anaesthesia has the most mature clinical AI of the five specialties here. Hypotension prediction software and ultrasound nerve-block guidance are cleared for use in the US. Sepsis and deterioration prediction in the ICU are widely studied, but benefit in randomised trials is mixed.
What AI does in Anaesthesiology and Critical Care
Software warns the anaesthetist that blood pressure is likely to fall within about 15 minutes.
Examples: Edwards Acumen HPI
AI labels nerves and nearby structures on the live ultrasound image.
Examples: ScanNav Anatomy Peripheral Nerve Block
Combines labs and clinical data to estimate sepsis risk in hospital and emergency patients.
Examples: Prenosis Sepsis ImmunoScore
Drafting handovers and answering clinical queries. Accuracy needs checking every time.
Examples: ChatGPT, OpenEvidence
Models trained on ICU records predict events such as shock or mortality.
Examples: Models built on MIMIC-IV
AI helps acquire and read lung, heart and vascular ultrasound views.
Examples: Handheld ultrasound AI tools
routine = in everyday use in many centres · early adoption = in use at some sites, evidence still building · research = studies only
Key ideas for this specialty
New to the basics? Start with the theory lessons.
Tools
Predicts hypotension from the arterial waveform; works with invasive and finger-cuff monitoring.
Regulatory status: FDA De Novo (DEN160044, 2018); later FDA 510(k) for finger-cuff use
Real-time AI labelling of anatomy on ultrasound for nine common blocks.
Regulatory status: FDA De Novo (October 2023)
AI sepsis risk score using 22 parameters, shown in the electronic record.
Regulatory status: FDA De Novo (April 2024)
Open dataset of intraoperative vital signs from 6,388 surgical cases.
Regulatory status: Not listed
Large open ICU database used to build critical care models. Needs credentialed access.
Regulatory status: Not listed
A listing here is not an endorsement. Ask any vendor for its CDSCO licence and class before clinical use.
Where to learn
Three courses on image diagnosis, prognosis and treatment models. Needs basic Python.
No-maths introduction to what AI can and cannot do. Good first step.
India's own ethics framework. Useful before you plan a thesis using AI.
Talks and symposium recordings on medical AI.
Two podcast episodes with guided reflection, 1.5 CME. Anaesthesia-specific.
Machine learning basics written for anaesthesiologists.
Do it yourself
Download a few cases from VitalDB. Plot arterial pressure and mark every mean pressure under 65 mmHg for one minute. Count events per case.
Open the dataset or tool →Complete the PhysioNet credentialing. Pull first-day ICU labs and fit a simple logistic model for in-hospital death. Compare it with SOFA.
Open the dataset or tool →Use public or properly de-identified data only, and get ethics approval before any project on your own patients.
Papers worth reading
- Predictive ability of hypotension prediction index and machine learning methods in intraoperative hypotension: a systematic review and meta-analysis Journal of Translational Medicine, 2024
- The applications of ChatGPT and other large language models in anesthesiology and critical care: a scoping review Canadian Journal of Anesthesia, 2025
- Artificial intelligence and machine learning in critical care research Journal of Critical Care, 2024
- Implementing Large Language Models in Critical Care CHEST Critical Care, 2025
Societies and guidance
Runs AI education such as AI from A to Z.
Critical care education and data science content.
National body; no AI guidance page confirmed.
Links checked 3 October 2026. “Not listed” means we could not confirm the detail, not that it is absent. Educational material, not clinical advice.