Construction of ALS Phenotype Knowledge Graph and Clinical Decision Support Based on MIMIC-IV

Amyotrophic Lateral Sclerosis (ALS) is a rapidly progressive neurodegenerative disease characterized by significant phenotypic heterogeneity, which complicates early diagnosis and personalized treatment planning. The exponential growth of electronic health records (EHRs) offers a data-rich environment for characterizing disease progression; however, the unstructured nature of clinical notes and the high dimensionality of structured data present substantial challenges. This paper proposes a comprehensive framework for constructing an ALS Phenotype Knowledge Graph utilizing the Medical Information Mart for Intensive Care IV (MIMIC-IV) database. We define a multi-layer ontology encompassing clinical symptoms, laboratory markers, pharmacological interventions, and physiological signals specific to ALS patient trajectories in critical care settings. By employing named entity recognition and relation extraction techniques, we transform raw clinical data into structured semantic triples. Furthermore, we develop a clinical decision support mechanism based on graph reasoning and semantic similarity to assist in risk stratification and therapeutic recommendation. Our experimental results demonstrate that the proposed knowledge graph effectively captures complex associations between diverse clinical entities. The subsequent decision support module exhibits high potential in identifying atypical phenotypic patterns, thereby facilitating more precise clinical interventions.