Semiautomatic Segmentation of Ventilated Airspaces in Healthy and Asthmatic Subjects Using Hyperpolarized MRI
Figure 1
Detailed schematic of semiautomatic segmentation algorithm. The example shown here is from a healthy subject. The first step is a statistical noise subtraction to generate an initial binary mask of the input image (a). Thereafter, the resultant lung mask (b) is refined through a four-class FCM clustering which partitions the entire image into four categories: negligible ventilation, low ventilation, intermediate ventilation, and high ventilation (c). Pixels that fall within the negligible ventilation class are subsequently discarded to form a corrected mask (d). Through a semiautomatic trachea removal involving a seeded region-growing algorithm, an area filter for connectivity, and a series of morphological operations, a final binary image representative of ventilated airspaces is obtained (e).
