AI-Driven Privacy-Preserving Medical File Processing Using Large Language Models

Large language models (LLMs) have recently attracted attention in healthcare by demonstrating impressive data analysis, summarization, and decision-making capabilities. Medical records also contain sensitive information about patients, which calls for strong privacy-preserving techniques. This paper presents a real-time privacy-preserving AI framework that applies CRP-based encapsulation, NLP-driven anonymization, and LLMs assisted summarization to secure medical file processing. Before LLMs can access data, NLP-based anonymization techniques are used to ensure that the data are compliant with privacy regulations. It provides patient anonymization, summarized medical opinions, and AI-driven recommendations, with no personal identifiable information (PII). We established an exemplary experimental performance case for medical text analysis using GPT-4 enhanced by encrypted, anonymized, and encapsulated CRP-encrypted records.