Reading time: approximately 10 minutes
TL;DR: A 2026 narrative review in Medical Gas Research identifies four areas where AI is reshaping hyperbaric oxygen therapy: protocol optimisation, patient selection, biosensor monitoring, and predictive safety analytics. For Canadian clinic owners, the question is where to start.
Artificial intelligence in hyperbaric oxygen therapy is the use of machine learning, digital-twin modelling, and connected biosensors to personalise treatment protocols, refine patient selection, monitor real-time physiology, and predict adverse events. The field has been discussed in academic circles for years, but the volume of peer-reviewed evidence has accelerated sharply since 2022. For a Canadian clinic owner planning capital purchases, accreditation milestones, or staff training for the next operational cycle, the 2026 evidence base is now substantive enough to inform strategy.
This outlook synthesises the most recent peer-reviewed literature and translates it into operational implications for clinic owners across Canada. Canada Hyperbarics maintains a vendor-neutral position on technology and accreditation pathways. Our role is to help owners, operators, and patients understand the evidence base and make informed decisions.
Why are Canadian clinic owners watching AI in hyperbaric medicine now?
Three pressures are converging on Canadian hyperbaric clinics at the same time. First, the indications base for hyperbaric oxygen therapy continues to broaden, with growing interest in neurological and inflammatory conditions alongside the established treatments for decompression sickness, carbon monoxide poisoning, radiation injury, and diabetic foot ulcers. Second, outcomes across centres remain constrained by heterogeneous protocols, limited patient selection tools, and the absence of real-time physiological guidance during sessions. Third, regulators and accreditation bodies are paying closer attention to safety data, dose standardisation, and adverse-event reporting.
These are exactly the operational gaps that AI tools have been designed to address in other clinical fields. The most cited 2026 narrative review on the topic, published by Epelde in Medical Gas Research, makes the case that machine learning, explainable AI, and digital twins could enable a precision approach to hyperbaric medicine, mirroring shifts already underway in oncology, cardiology, and critical care (read the full study summary).
For owners, the strategic point is not that AI will replace hyperbaric technologists or supervising physicians. It is that the operational layer around HBOT delivery, the layer responsible for scheduling, patient selection, dose calibration, sensor monitoring, and incident detection, is becoming a software-augmented layer. Clinics that begin understanding this layer in 2026 will be better positioned for the accreditation expectations of 2028 and beyond.
What does the latest 2026 research say about AI integration with HBOT?
The Epelde 2026 review searched PubMed, Scopus, Web of Science, and Google Scholar across the January 2000 to March 2025 window using terms combining hyperbaric oxygen therapy with artificial intelligence, machine learning, deep learning, digital twins, and biosensors. Fifty-three eligible studies met inclusion criteria, with at least 30 per cent published between 2022 and 2025. The review groups the literature into four convergent application areas. Each maps directly to a clinic-operations function.
Protocol optimisation using multicentre data. AI models trained on large, harmonised registries can identify the protocol parameters most associated with favourable outcomes for specific indications. This matters because the strongest current evidence for real-world heterogeneity comes from registries like the Italian ITA-OTI multicentre prospective observational study, which enrolled 327 patients across 10 centres in a 12-week window. That registry found that treatments clustered at a median of 2.5 atmospheres absolute with two oxygen cycles per session, but with notable heterogeneity in the number of sessions per patient (Ippolito et al., 2026). The same kind of data, collected at the Canadian level, could feed the protocol-optimisation models Epelde describes.
Biomarker-driven patient selection. Multi-omics profiles, imaging features, and laboratory markers can be combined by machine-learning classifiers to predict who is most likely to benefit from HBOT for a given indication. This is potentially the highest-leverage use case for clinics, because patient selection drives both clinical outcomes and operational efficiency.
Real-time adaptive control using biosensors. Wearable and inline sensors can stream physiological data during a hyperbaric session, allowing for adaptive adjustments rather than fixed protocols. A small 2026 prospective cohort study in Comprehensive Physiology used portable single-lead ambulatory ECG devices to track heart rate variability across five daily 90-minute HBO sessions in 14 young adults, with measurements taken at baseline, two hours after the fifth session, and again one month later. The investigators reported significant changes in autonomic nervous system indices over time, demonstrating that wearable monitoring during and after HBOT generates clinically interpretable signals (Zhang et al., 2026).
Predictive safety analytics for oxygen toxicity and barotrauma. Risk-scoring models trained on adverse-event registries can flag patients at elevated risk before sessions begin. This category is the most directly relevant to incident reporting, accreditation reviews, and Health Canada inspection preparation.
Which AI use cases matter most to clinic operations?
Not every application area is equally mature, and not every use case is equally relevant to a clinic owner setting a 2026 capital budget. The table below maps the four application areas described in the Epelde review against operational relevance, capital intensity, and realistic 2026 deployment maturity for a Canadian clinic.
| Application area | Operational function affected | Capital intensity | 2026 deployment maturity |
|---|---|---|---|
| Protocol optimisation from multicentre data | Clinical governance, protocol committees | Low (data sharing, no new equipment) | Early, dependent on registry participation |
| Biomarker-driven patient selection | Intake screening, referral triage | Low to moderate (lab and imaging integration) | Pilot stage, mostly research settings |
| Real-time adaptive control via biosensors | In-session monitoring, technologist workflow | Moderate (wearable sensors, software) | Demonstration evidence, not yet routine |
| Predictive safety analytics | Incident reporting, accreditation, insurance | Low (software layered on existing records) | Most actionable for 2026 to 2027 |
For most owners, the lowest-cost and highest-impact place to start is predictive safety analytics built on existing incident and adverse-event records. The infrastructure already exists in any clinic that meets current documentation standards. The next priority is participation in or alignment with multicentre data initiatives, because protocol-optimisation models are only as good as the data feeding them, and Canadian data representation matters for Canadian operational decisions.
How might real-time biosensors change session delivery?
Biosensor integration is the application area where clinic owners will feel the most concrete operational change. Two 2026 studies illustrate the opportunity and the current limits.
On the opportunity side, the Zhang heart rate variability cohort showed that consumer-grade portable single-lead ambulatory ECG devices can capture meaningful autonomic data through a structured HBOT course. Five daily 90-minute sessions produced measurable shifts in sympathetic tone-related indices at one month, with no significant changes in vagal indicators and a reduced maximum heart rate at follow-up. This is a small study and the cohort was 14 young male participants, so the findings are not yet generalisable to a typical clinic population. The methodological point still stands: wearable physiological monitoring around HBOT sessions is technically feasible with off-the-shelf equipment.
On the limits side, a 2026 prospective observational study in Scientific Reports evaluated non-invasive pulse CO-oximetry against arterial carboxyhaemoglobin in 81 patients with acute carbon monoxide poisoning. At baseline, the non-invasive measurement showed only limited agreement with the arterial reference, with moderate concordance in the moderate severity range but substantial positive bias and wide limits of agreement at higher concentrations. After HBOT, agreement between the two measurements deteriorated markedly (Lee et al., 2026). The clinical implication is that current generations of some non-invasive sensors are not yet reliable replacements for arterial measurement in critical settings. The operational implication is that biosensor adoption in 2026 should be additive, not substitutive, and clinics should be cautious about pricing or marketing claims that overstate sensor accuracy.
Computational modelling is also informing what protocols might look like under sensor-driven adaptive control. A 2026 paper in PLoS One developed partial differential equation and agent-based simulation models of symmetric flat ischaemic dermal wounds, including the role of keratinocytes, which make up around 90 per cent of the cells in the epidermis. The two model approaches showed high agreement, and both suggested that standard hyperbaric and topical oxygen therapies effectively close wounds in expected time when ischaemia is not too severe (Lazebnik et al., 2026). These models are the kind of computational scaffolding that digital-twin systems for clinical use will rest on.
What practical steps can Canadian clinic owners take in 2026?
Operationalising AI integration without overcommitting is the realistic 2026 posture for most Canadian clinic owners. The following five steps are derived from the application areas in the Epelde review and grounded in what current Canadian clinic infrastructure can support.
- Audit your incident and adverse-event records. Predictive safety analytics start with structured historical data. Make sure your records are coded consistently, timestamped, and linkable to patient indication, session count, and pressure protocol.
- Standardise session-level documentation. Median pressure, oxygen cycles per session, total session count by indication, and pre-session screening fields are the same data points the Italian ITA-OTI registry tracked. Your future ability to participate in Canadian registries depends on that data already existing in usable form.
- Pilot one biosensor stream cautiously. Heart rate variability monitoring via portable single-lead ambulatory ECG is the most evidence-supported entry point. Treat it as research-adjacent until larger studies generalise the findings to a clinic population.
- Engage with national associations. The Canadian Undersea and Hyperbaric Medical Association is the most direct vehicle for shaping Canadian data initiatives and shared protocol committees. Owners who participate now influence the registry design that AI models will eventually train on.
- Hold the line on regulated practice. AI software that supports clinical decisions falls under medical device regulatory frameworks. Any deployment of a clinical decision-support tool in a Canadian hyperbaric setting needs to be evaluated against Health Canada device classification rules and your accreditation body’s expectations.
What are the regulatory and accreditation considerations?
Hyperbaric chambers in Canada are regulated by Health Canada as medical devices. Software that interprets sensor data and informs clinical decisions can also be regulated as a medical device, depending on its function and risk classification. Owners should consult both Health Canada guidance on medical device software and the Undersea and Hyperbaric Medical Society accreditation framework before deploying any AI-enabled clinical tool in routine practice.
Provincial public coverage for HBOT remains the dominant reimbursement reality. Owners can review the current state of provincial coverage on our coverage guide, which is updated as provincial plans, accreditation bodies, and indications change. Where clinics treat patients across provincial lines, the regulatory and reimbursement environment differs by jurisdiction, and AI-enabled protocols do not change that underlying landscape.
Patients searching for accredited Canadian hyperbaric care can browse our directory of hospitals and regulated facilities at canadahyperbarics.ca/facilities/, which lists hospital-based programmes and accredited private clinics across Canadian provinces.
Frequently asked questions
Is AI in hyperbaric oxygen therapy a regulated medical device in Canada?
Software intended for clinical decision support can fall under Health Canada’s medical device framework, depending on its function and risk class. Owners deploying AI-enabled tools should consult Health Canada guidance and confirm classification before clinical use.
What is the most realistic 2026 entry point for AI integration at a Canadian HBOT clinic?
Predictive safety analytics built on existing incident and adverse-event records is the lowest-cost starting point. The infrastructure usually already exists, and the application area aligns directly with accreditation and Health Canada inspection expectations.
Are wearable biosensors accurate enough to replace arterial measurements during HBOT?
Not in all settings. A 2026 prospective study of non-invasive pulse CO-oximetry against arterial carboxyhaemoglobin in acute carbon monoxide poisoning found limited agreement at baseline and worsening agreement after HBOT. Current biosensors should be treated as additive monitoring tools, not substitutes for arterial measurement in critical scenarios.
Do digital twins for hyperbaric oxygen therapy exist clinically today?
Computational models of HBOT-relevant physiology are emerging in the research literature, including partial differential equation and agent-based simulations of ischaemic wound closure. Clinical-grade digital twins are not yet in routine practice, but the computational scaffolding is forming.
What should a Canadian clinic owner do to prepare for AI-driven protocols?
Standardise documentation at the session level (pressure, oxygen cycles, session count by indication), audit incident records, engage with the Canadian Undersea and Hyperbaric Medical Association on national data initiatives, and pilot biosensor monitoring in a research-adjacent posture before clinical adoption.
Where can patients find accredited HBOT facilities in Canada?
Canada Hyperbarics maintains a directory of hospitals and regulated facilities providing hyperbaric oxygen therapy across Canadian provinces. The directory is updated as new facilities open, accreditation status changes, or programmes are added.
What comes next for AI and hyperbaric medicine in Canada?
The 2026 evidence base is best read as a signal that the field has moved past pure speculation. Fifty-three peer-reviewed studies linking HBOT with AI, machine learning, digital twins, or biosensors is a body of work. It is not yet a body of practice. The gap between the literature and routine Canadian clinic operations is where the next two to three years will be decided.
For most Canadian clinic owners, the right 2026 posture is to read carefully, document thoroughly, engage with national associations, and pilot the lowest-risk applications. The owners who treat 2026 as a documentation and data-quality year will be the ones best positioned when accreditation bodies and regulators formalise expectations around AI-enabled hyperbaric practice. Canada Hyperbarics will continue tracking peer-reviewed publications and translating them into practical implications for Canadian operators.
To explore the underlying research bank, browse our research database of indexed HBOT studies. To find an accredited Canadian programme, browse our directory of hospitals and regulated facilities.
This content is for informational purposes only and does not constitute medical advice. Hyperbaric oxygen therapy decisions should be made in consultation with a qualified physician. Regulatory and accreditation requirements change over time; verify current requirements with Health Canada and your accreditation body before making operational decisions.