Artificial intelligence in quantitative chest imaging analysis for occupational lung disease: appraisal of its current status.

purpose of reviewThe application of mathematical algorithms for detecting lung abnormalities has been a challenge for decades. Occupational lung diseases, which often present as diffuse abnormalities, are primarily screened and diagnosed using chest radiographs and computed tomography (CT). This article reviews recent algorithmic advancements applied to these diagnostic tasks. RECENT

findingsSignificant progress has been made in artificial intelligence (AI) technologies, particularly with three-dimensional deep learning models based on convolutional neural networks (CNNs). For chest radiographs, promising approaches include the "eTóraxLaboral" platform for pneumoconiosis detection, CNNs enhanced with dark channel prior-inspired lesion area enhancement, and CNNs paired with CycleGAN. For CT imaging, transformer-based factorized encoders (TBFE), various CNN architectures (often combined with other techniques), and the recently developed Kolmogorov-Arnold Networks (KANs) for binary classification have shown strong performance. However, both chest radiograph and CT studies commonly rely on the International Labour Organization (ILO) International Classification of Radiographs of Pneumoconioses system (ILO/ICRP) for pneumoconiosis as a reference, which may limit AI development for CT in particular. SUMMARY: Recent advancements offer strong promise for computer-assisted diagnosis of pneumoconiosis using chest radiographs and CT scans. The standardization and integration of these technologies - especially with support from international organizations and collaborative studies - will be critical to achieving accurate, implementable screening tools for occupational lung disease.