Artificial intelligence-based clinical decision support systems in neurophysiology: conceptual framework, design criteria, and ethical considerations

Duliet Hong León; Jaime Perez Fernández; Jaime Azpiazu Saiz

Resumen:

Diagnosis in neurophysiology using visual evoked potentials (VEP) for multiple sclerosis and optic neuropathies remains specialist-dependent and subject to variability. Artificial intelligence (AI) offers transformative potential, but its responsible implementation is hindered by algorithmic metrics. This paper develops a conceptual framework for AI-based clinical decision support systems (CDSS) in neurophysiology, structured around three integrated dimensions: CDSS typology and foundations, responsible design criteria, and ethical-regulatory considerations. Six architectural principles are proposed: transparency, clinical anchoring in validated criteria (ISCEV 2023), preservation of physician judgment, false-alarm minimization, reproducibility, and interoperability, applicable to any system in the field. The framework is illustrated through an EEG‑VEP system on the VEPCON dataset, integrating automated classification, clinical hierarchy over statistical thresholds, and a dashboard with an explicit ethical disclaimer. The responsible development of CDSS in neurophysiology demands a systemic approach that integrates clinical evidence, operational design, and medical device regulation, ensuring their role as support rather than as a substitute for clinical judgment.