Research progress on artificial intelligence driving precision diagnosis and treatment of chronic obstructive pulmonary disease
[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.
chronic obstructive pulmonary diseaseartificial intelligenceunimodal datamultimodal dataartificial general intelligencegenerative artificial intelligencemultimodal large language modelrespiratory medicineprecision medicine
国家科技期刊平台
NSTL
万方数据
