人工智能驱动慢性阻塞性肺疾病精准诊疗研究进展

Research progress on artificial intelligence driving precision diagnosis and treatment of chronic obstructive pulmonary disease

[背景]慢性阻塞性肺疾病(chronic obstructive pulmonary disease,COPD)是一种全球常见的慢性呼吸系统疾病,其早期精准诊断和治疗对患者生活质量有着重大影响.近年来,人工智能(artificial intelligence,AI)技术在医疗领域的快速发展,为COPD早期精准诊疗开辟了新的思路.[进展]本文梳理了AI技术在COPD诊疗中的应用现状,尤其是AI单模态和多模态模型的应用研究进展.单模态模型专注于单一类型的数据源,显示了其早期诊断和监测能力;而多模态模型通过融合来自医学影像、生物医学数据、电子病历等多源信息,进一步提升了对COPD患者病情的全面理解和个性化精准治疗,具有更加广阔的应用前景.[展望]AI技术在COPD早期诊断、区分疾病严重程度、预测急性加重、治疗、管理监测及康复等方面展示出独特的优势.尤其是当前通用AI、生成式AI以及多模态大语言模型等前沿AI技术的快速发展,必将大力促进医生更加精确地诊断疾病,制定更加个性化的患者治疗方案,大幅度地提高临床治疗效果和患者生活质量.

[Background]Chronic obstructive pulmonary disease(COPD)is a complex and prevalent respiratory disorder with irreversible airflow limitation worldwide.Precision diagnosis and treatment at its early stage significantly improve the quality of life of patients.COPD symptoms are diverse and progressive,e.g.,chronic cough,sputum production,dyspnea and chest tightness,indicating advances in COPD.While the pathophysiology of COPD is multifaceted with persistent airway inflammation,airway remodeling,and alveolar destruction,the etiology of COPD is multifactorial,including prolonged smoking,environmental pollutants,occupational hazards,and genetic predispositions.These factors collectively result in airflow obstruction and pathological changes in the respiratory tract.Specifically,the progression of COPD is often accompanied with persistent inflammatory responses,oxidative stress,and intensive pulmonary damage.[Progress]Pulmonary function tests(PFTs)are routinely performed to examine COPD,providing physicians with a ratio of the forced expiratory volume in one second by the forced vital capacity to evaluate COPD.Unfortunately,the results of PFTs are critically affected by the effort of patients,and the interpretation of PFTs also depends on experience and skills of physicians.While PFTs allow physicians to quantify the severity of COPD,they do not reach a specific diagnosis and are commonly associated with medical history,physical examination such as CT imaging,functional MR imaging and respiratory sound,and laboratory data to determine a diagnosis.Therefore,physicians expect more precise COPD diagnosis and treatment methods than conventional ones to improve patient s quality of life.Nowadays artificial intelligence(AI)is widely discussed in precision medicine.Specifically,AI techniques or mathematical models are also increasingly used in COPD diagnosis,treatment,monitoring,and management.These models are generally categorized into unimodal and multimodal AI models in accordance with clinical COPD data.While the unimodal model uses only a single one modality such as PFTs or CT images,the multimodal model fuses a diversity of data including imaging,biomedical information,and clinical records.All these models generally provide physicians with a holistic assessment of COPD,patient-specific treatment for precision medicine.[Perspective]In general,AI techniques provide a promising way to precisely diagnose and treat COPD in its early stage,as well as COPD management and monitoring.Specifically,artificial general intelligence,generative artificial intelligence,multimodal large language models are innovating clinical methods in diagnosis,treatment,monitoring,and management of pulmonary diseases,although they still suffer from medical data privacy and security,model generalizability,interpretability and complexity,legal and ethical issues.Future research should address these issues in various angles.It is essential to strengthen privacy protection and security measures.Moreover,it is vital to improve the generalizability,transparency and interpretability and reduce the complexity of various AI models in clinical applications.Additionally,medical ethics are important when applying AI techniques to precision pulmonary medicine.