Radiologist-inspired Symmetric Local–Global Multi-Supervised Learning for early diagnosis of pneumoconiosis

Abstract

Pneumoconiosis is a severe occupational lung disease caused by long-term exposure to inhaled dust, where early diagnosis is critical for effective management and health protection. However, current deep learning approaches struggle with the subtle radiographic manifestations of pneumoconiosis, strict diagnostic criteria, and limited data availability. In this paper, we propose Symmetric Local–Global Multi-Supervised Learning (SLGMS), a novel framework inspired by the diagnostic practices of specialized radiologists. SLGMS integrates a mechanism for generating symmetric global and local views with a symmetric VMamba feature extraction network, effectively mimicking the region-by-region analysis and comparative assessment of symmetric regions performed by radiologists. Additionally, it incorporates a local–global knowledge distillation architecture with tailored multi-supervised learning to explore relationships between local and global views while adhering to clinical diagnostic criteria for pneumoconiosis. Evaluated on pneumoconiosis datasets collected from two medical hospitals in China, SLGMS demonstrates superior performance, achieving an average improvement of 6.19% in accuracy, sensitivity, specificity, and AUC metrics on the internal test set and 3.28% on the external validation dataset compared to state-of-the-art methods. On the public NIH ChestX-ray14 benchmark, a transferable variant of SLGMS achieved a new state-of-the-art AUC of 82.9%, while the full SLGMS provides an average improvement of 3.5% on its supplemental fibrosis dataset. By bridging diagnostic prior knowledge with deep learning, SLGMS offers an effective paradigm for early diagnosis of occupational pneumoconiosis in data-scarce environments, with broader applicability and scalability to other thoracic diseases.