Advances in AI for detecting pulmonary inflammation and perioperative medicine: a mini-review.
Authors
Affiliations (2)
Affiliations (2)
- Department of Anesthesiology, Guangxi Hospital Division of The First Affiliated Hospital, Sun Yat-sen University, Nanning, Guangxi, China.
- Department of Anesthesiology, The First Affiliated Hospital, Sun Yat-sen University, Guangzhou, Guangdong, China.
Abstract
With increasing human longevity, early recognition and treatment of pneumonia in the elderly are crucial to prevent disease progression. Artificial intelligence (AI) is rapidly transforming the detection and management of pulmonary inflammation (pneumonia, COVID-19 lung damage). Accurate preoperative assessment of pneumonia contributes to improved perioperative surgical and anesthesia management. This mini-review highlights key advances: (1) Hybrid deep learning models achieve high accuracy (>96%) in analyzing ultrasound videos for disease differentiation. (2) Self-supervised learning enables expert-level X-ray interpretation without extensive annotations. (3) Multimodal integration combines imaging (CT/X-ray) with clinical data, enhancing lesion visibility and pathogen-specific diagnosis (viral vs. bacterial AUC: 0.95). Clinically, AI demonstrates high efficacy in COVID-19 detection (AUC: 0.992), pediatric pneumonia diagnosis (89-96% accuracy), and identifying post-COVID complications. Despite this promise, challenges remain, including data bias, limited pediatric datasets, "black-box" model interpretability, and ethical concerns. Future progress depends on expanding diverse training data (e.g., via federated learning), integrating explainable AI (XAI), and ensuring equitable access. In conclusion, AI offers accurate, scalable solutions for pulmonary inflammation diagnostics, with significant potential to augment clinical decision-making and extend into proactive areas like perioperative medicine for complication screening and prevention.