Published August 13, 2025
| Version v1
Preprint
Open
Symptom Cascade Analyzer: A Graph-Theoretic Natural Language Processing Framework for Culturally-Adaptive Medical Diagnosis
Description
The manuscript introduces a graph-theoretic foundation for modeling medical knowledge, incorporating cultural glossaries, multilingual embeddings based on XLM-RoBERTa, bias adjustment mechanisms, and rare disease detection via cluster entropy.
It details the SCA algorithm, experimental results on a multilingual corpus of 15,000 symptom-disease pairs, and ablation studies showing incremental performance gains.
Future directions include expansion to additional dialects, pilot studies in rural African clinics, federated learning for privacy, and IRB-approved clinical validation.