An NLP-Based Neuro-Semantic Clinical Filter for Medical Text Simplification

Medical texts are often complex and difficult to understand for non-specialists, creating barriers to effective communication in the clinical and rehabilitation fields. Although recent advances in natural language processing (NLP) have enabled automated text simplification, existing approaches often struggle to maintain medical accuracy and frequently result in factual inconsistencies or distortions. To address these issues, we propose the Neuro-Semantic Clinical Filter (NSCF), a novel NLP-based framework designed for clinically accurate simplification of medical texts. The proposed method integrates a Medical Concept Graph Encoder (MCGE) to incorporate structured domain knowledge, a Neuro-Symbolic Transformer (NSTR) for supervised text generation, and a Knowledge Integrity Validator (KIV) to ensure factual consistency during decoding. Furthermore, an adaptive module (ClinAdapt) enables text personalization based on patient profiles and comprehension levels. Extensive experiments were conducted on several medical text corpora, comparing NSCF with modern baseline models, including BART, T5, and domain-specific variants. Experimental results show that the NSCF demonstrates improved performance across several metrics, including a SARI score of 47.2, a BERTS score of 91.6, and a factual alignment ratio (FAR) of 88.1. Human evaluation further confirms improvements in reading fluency, accuracy, and clinical applicability. These results highlight the effectiveness of the proposed approach in bridging the gap between complex medical language and patient-level understanding, providing a robust and interpretable solution for real-world healthcare applications.