A hybrid multi-scale approach to automatic airway tree segmentation from CT scans

Airway structure and morphology is commonly related to inflammatory and infectious lung diseases, and often analyzed non-invasively through high resolution computed tomography (CT) scans. Conventionally, most airway related feature characterization on these scans is performed manually, but is often too labor intensive and time consuming for routine clinical practice. Therefore, semi- and fully-automatic airway segmentation algorithms are crucial for the diagnosis of these conditions. A fundamental challenge in airway tree segmentation is highly variable intensity levels within the lumen, which often causes a segmentation method to leak into adjacent lung parenchyma through blurred airway walls or soft boundaries. In this paper, we present a new hybrid multi-scale airway segmentation approach to solve these problems through proposing a new fuzzy connectivity based algorithm combining multiple features to identify airways at different scales. The performance of the proposed method was qualitatively and quantitatively evaluated on pulmonary CT images from human patients with diverse diseases with promising results.