Artificial intelligence and machine learning in thoracic surgery- A scoping review

Abstract

Artificial intelligence (AI) and machine learning (ML) are rapidly transforming thoracic surgery, offering innovative solutions to enhance patient care, improve surgical outcomes, improve surgical training, and increase efficiency. This scoping review provides a comprehensive overview of the current applications, challenges, and future directions of AI and ML in thoracic surgery. Key applications of AI in thoracic imaging include lung nodule detection and characterisation, with deep learning algorithms demonstrating performance comparable to or exceeding that of human radiologists. Radiomics combined with ML techniques show promise in tumour characterisation and classification of non-small cell lung cancer subtypes. In preoperative planning, AI-powered 3D reconstruction and virtual reality systems enable detailed surgical simulation and risk assessment. Augmented reality and computer-assisted navigation systems are being developed to enhance surgical precision intraoperatively. While fully autonomous robotic surgery remains a distant goal, AI-enhanced robotic platforms are advancing rapidly. Postoperatively, AI algorithms show potential for predicting outcomes, interpreting pulmonary function tests, and guiding rehabilitation strategies. Despite these advancements, several challenges persist, including data quality and quantity issues, algorithm interpretability, and the need for rigorous clinical validation. Ethical considerations surrounding AI implementation in healthcare also require careful attention. Future directions include integrating multimodal data, developing real-time intraoperative guidance systems, and creating adaptive AI models capable of continuous learning. As these technologies mature, they have the potential to revolutionise thoracic surgical practice, ultimately improving patient outcomes. • AI algorithms for lung nodule detection and characterisation demonstrate performance comparable to or exceeding that of human radiologists. • Radiomics combined with ML techniques shows promise in non-invasive tumour characterisation and classification of non-small cell lung cancer subtypes. • AI-powered 3D reconstruction and virtual reality systems enable detailed surgical simulation and risk assessment for thoracic procedures. • Deep learning models like MesoNet can predict overall survival in mesothelioma patients from histology images, outperforming traditional classification methods. • AI-enhanced robotic platforms and augmented reality systems are advancing rapidly, showing potential to improve surgical precision and outcomes in thoracic surgery.

Ähnliche Arbeiten