Leveraging artificial intelligence for autonomous space healthcare: current evidence, operational architecture and research priorities for deep-space exploration

Introduction Deep-space missions will progressively limit real-time ground support, rapid evacuation and resupply, creating a need for medical capabilities that can operate with graded autonomy. Artificial intelligence (AI) may support this transition, but terrestrial performance cannot be equated with spaceflight readiness. Methods We conducted a narrative, hypothesis-generating review of literature relevant to AI-enabled space healthcare. Sources were selected iteratively for their contribution to health monitoring, biofluid analysis, medical imaging, clinical decision support or radiation assessment. The review did not use a systematic screening protocol or formal meta-analytic quality assessment; therefore, findings are synthesised qualitatively and claims are calibrated to evidence maturity. Results Evidence was grouped as spaceflight operational use, space-relevant prototype or analogue demonstration, terrestrial clinical validation, terrestrial experimental demonstration, or conceptual transfer. AI may support personalised monitoring, biofluid interpretation, ultrasound acquisition and analysis, radiation forecasting and clinical decision support. However, most cited applications remain terrestrial demonstrations. Key barriers include limited astronaut data, shifted physiological baselines, restricted computation and power, sensor and storage constraints, model drift, explainability, uncertainty management and failure handling. Conclusion AI is most plausibly deployed as one component of an integrated autonomous medical system that combines wearables and biosensors, onboard records, imaging, radiation monitoring, clinical decision support, crew oversight and delayed telemedical consultation. Near-term priorities are prospective validation in mission-relevant analogues, personalised baselines, resource-aware edge deployment, explicit uncertainty thresholds, human override and pre-defined safe fallback procedures. AI should augment rather than replace telemedicine and professional judgement.