Neuro-fuzzy systems in internet of medical things: a systematic review on applications, taxonomy, challenges and open issues
Résumé
The growing demand for real-time, adaptive, and explainable analytics in the Internet of Medical Things (IoMT) has increased interest in neuro-fuzzy systems for connected healthcare. By combining neural learning with fuzzy inference, these systems can support predictive modelling while offering rule-based reasoning structures that require explicit evaluation for interpretability, stability, and clinical usability. This paper presents a systematic review based on structured database searches, eligibility screening, quality appraisal, and narrative synthesis of 55 peer-reviewed journal articles published between January 2020 and July 2025 on neuro-fuzzy systems in IoMT and connected-health contexts. The included articles were retrieved from MDPI, SpringerLink, ScienceDirect/Elsevier, IEEE Xplore, Wiley Online Library, and Taylor & Francis, with Google Scholar used only for verification and citation tracing. An article-level, deployment-aware taxonomy was developed to classify the evidence by architecture, application, design, deployment environment, and reported metrics. Neural-network-based optimization was the most frequently reported architecture, accounting for 28 articles (50.9%), followed by hybrid neuro-fuzzy systems with 11 articles (20.0%), deep neuro-fuzzy systems with 9 articles (16.4%), and evolving neuro-fuzzy systems with 7 articles (12.7%). This distribution indicates that the evidence base is still shaped mainly by static or offline neuro-fuzzy designs, while evolving architectures for streaming, non-stationary, and patient-specific IoMT data remain comparatively underexplored. Predictive systems formed the largest application category, followed by detection-oriented systems. Deployment reporting remained limited, with 34 of the 55 included articles (61.8%) not specifying an execution environment, while only 6 articles (10.9%) reported latency or processing-time evidence. Overall, the evidence remains mainly retrospective, experimental, simulation-based, benchmark-driven, or prototype-level. The review identifies recurring gaps in online adaptability, interoperability, deployment-aware performance evaluation, and practical clinical interpretability. Although neuro-fuzzy systems show promise as adaptive models for connected-health applications, the current evidence does not yet establish routine clinical readiness or clinical effectiveness. Future work should strengthen evolving neuro-fuzzy modelling, prospective validation, deployment-aware evaluation, safety assessment, interpretability assessment, and integration with electronic health record and telemedicine workflows.
Citer ce document
Accès au document
Texte intégral en lecture en ligne, réservé aux abonnés SPHAERO et aux membres de l'institution. Se connecter
Voir l'article sur le site de la revueAuteur(s)
Statistiques
Consultations : 1
Téléchargements : 0