Frontiers in digital health · 2026
Explainable multimodal AI and neuro-symbolic clinical decision support system for chronic eye disease management: a digital health implementation study.
СтатьяНазвание и аннотация: Перевод в очереди
Обработка статьи · этапы, ошибки и управление
Загрузка состояния…
Ошибки и ограничения
История · было → стало
Состояние обновляется каждые 5 секунд, пока блок открыт. Процент показывается только там, где известен объём работы. Пауза останавливает продолжение этапа; текущий запрос к источнику может завершиться.
Wang Mini Han, Lee Simon Ming Yuen, Hou Guanghui, Wang Yapeng, Alves José C, Xie Ruitao, He Yaqing, Liu Jin, Fang Xiaoxiao, Yang Yu, Cai Xiaodong, Zheng Shuai, Yu Ziyang, Lin Ethan Zhiyuan, Cheang Chonin, Ian Kuok Kai, Qin Shuai
Аннотация
INTRODUCTION: Administrative burden and documentation workload are increasingly recognized as major contributors to healthcare costs and clinician burnout, particularly in chronic disease management such as age-related macular degeneration (AMD). These non-clinical tasks reduce time available for patient care and introduce inefficiencies in coding, billing, and compliance workflows. This study aimed to evaluate the technical feasibility and economic impact of an artificial intelligence-based system for automating administrative documentation in AMD care.
METHODS: A longitudinal proof-of-concept study was conducted in a tertiary ophthalmology network. A hybrid Neuro-Symbolic and Large Language Model (LLM) framework was developed to automate information extraction, clinical documentation structuring, and billing code validation. The system was applied to 24 de-identified unstructured clinical documents, including outpatient notes, operative reports, optical coherence tomography reports, and billing records. The processing pipeline included optical character recognition, knowledge-augmented LLM-based entity extraction, and neuro-symbolic rule-based validation to ensure clinical consistency, ICD-10/CPT coding accuracy, and reimbursement eligibility. Outputs were structured into standardized JSON clinical documentation formats.
RESULTS: The system achieved 98.3% accuracy in clinical entity extraction and 96.7% accuracy in administrative information extraction. Automated rule validation achieved 100% reimbursement compliance with no denied insurance claims. Mean documentation time per encounter decreased from 25.0 ± 5.0 min to 3.2 ± 1.1 min, representing an 88% reduction in documentation time. This resulted in an estimated labor cost saving of approximately 52 CNY (≈7 USD) per visit and projected annual savings of 42-56 USD per AMD patient under standard treatment schedules.
DISCUSSION: This study demonstrates the feasibility of integrating neuro-symbolic reasoning with large language models to automate administrative workflows in ophthalmology. The proposed system improves documentation efficiency, coding accuracy, and auditability while reducing hidden administrative costs. Although limited by a small sample size and single-center design, the framework provides a scalable and explainable architecture for AI-assisted administrative automation in digital health systems and chronic disease management.
CONCLUSION: AI-enabled administrative automation using a neuro-symbolic and LLM framework has the potential to significantly improve operational efficiency and reduce costs in AMD care, supporting the development of sustainable and value-based digital healthcare systems.
METHODS: A longitudinal proof-of-concept study was conducted in a tertiary ophthalmology network. A hybrid Neuro-Symbolic and Large Language Model (LLM) framework was developed to automate information extraction, clinical documentation structuring, and billing code validation. The system was applied to 24 de-identified unstructured clinical documents, including outpatient notes, operative reports, optical coherence tomography reports, and billing records. The processing pipeline included optical character recognition, knowledge-augmented LLM-based entity extraction, and neuro-symbolic rule-based validation to ensure clinical consistency, ICD-10/CPT coding accuracy, and reimbursement eligibility. Outputs were structured into standardized JSON clinical documentation formats.
RESULTS: The system achieved 98.3% accuracy in clinical entity extraction and 96.7% accuracy in administrative information extraction. Automated rule validation achieved 100% reimbursement compliance with no denied insurance claims. Mean documentation time per encounter decreased from 25.0 ± 5.0 min to 3.2 ± 1.1 min, representing an 88% reduction in documentation time. This resulted in an estimated labor cost saving of approximately 52 CNY (≈7 USD) per visit and projected annual savings of 42-56 USD per AMD patient under standard treatment schedules.
DISCUSSION: This study demonstrates the feasibility of integrating neuro-symbolic reasoning with large language models to automate administrative workflows in ophthalmology. The proposed system improves documentation efficiency, coding accuracy, and auditability while reducing hidden administrative costs. Although limited by a small sample size and single-center design, the framework provides a scalable and explainable architecture for AI-assisted administrative automation in digital health systems and chronic disease management.
CONCLUSION: AI-enabled administrative automation using a neuro-symbolic and LLM framework has the potential to significantly improve operational efficiency and reduce costs in AMD care, supporting the development of sustainable and value-based digital healthcare systems.