Logo image
Back
Healthcare Solutions for Noisy Clinical Text: A Federated Privacy-Preserving Approach
Conference proceeding   Peer reviewed

Healthcare Solutions for Noisy Clinical Text: A Federated Privacy-Preserving Approach

A B M Kamrul Islam Riad, Abdullah Al Mamun, Salma Akter, Md Abdul Barek, Maliha Zaman Nizum, Guillermo Francia, Hossain Shahriar, Alfredo Cuzzocrea and Sheikh Iqbal Ahamed
IEEE International Conference on Big Data, pp.4360-4369
IEEE International Conference on Big Data (BigData) (Macau, China, 12/08/2025–12/11/2025)
03/06/2026
DOI: https://doi.org/10.1109/BigData66926.2025.11400798

Metrics

1 Record Views

Abstract

The rapid expansion of electronic health records (EHRs) presents significant opportunities for integrating artificial intelligence (AI) into healthcare workflows. However, the unstructured and noisy nature of clinical text complicates accurate information extraction and automated decision support. This study proposes a federated, privacy-preserving AI framework for processing and denoising clinical text while ensuring strict data confidentiality. The system leverages advanced natural language processing (NLP) techniques, particularly domainspecific transformer models such as BioClinicalBERT, combined with robust noise reduction methods. Federated learning enables decentralized model training across healthcare institutions without transferring raw patient data. To strengthen privacy, the framework integrates secure multi-party computation and user-level differential privacy, ensuring compliance with HIPAA and GDPR standards. Experimental evaluations on benchmark datasets demonstrate high performance, achieving a Bilingual Evaluation Understudy (BLEU) score of 99% and a word error rate (WER) as low as 0.01%, indicating exceptional fluency and accuracy in clinical interpretation. This approach offers a scalable and secure solution for real-world clinical text analysis, supporting interoperability and collaboration across institutions. Future work will involve deployment in hospital settings to assess its impact on clinical workflows and decision-making, paving the way for more reliable AI-assisted healthcare applications.care outcomes.

Details

Logo image