Abstract—This paper presents a method to detect pulmonary boundary from an endobronchial ultrasonogram which can automatically assist bronchoscopists to define the region of interests within obtained endobronchial ultrasonograms images in order to perform further texture analysis aiming to diagnose the lesion. Multiple image processing techniques are performed to capture lesion boundary, starting from transforming image pixels from Cartesian coordinates into Polar coordinates, iterating over each degree to find pixels with the highest intensity, grouping candidate pixels with Density-Based Spatial Clustering of Applications with Noise (DBSCAN), eliminating isolate groups by second order derivatives, smoothing an image using moving average filtering, connecting remaining pixels to form a boundary using cubic spline interpolation. The total of 200 images, which composed of 100 benign and 100 malignant images, are used to evaluate the proposed method. As a result, the correlation between the obtained boundary and the ground-truth data is 93.92%.
 

Index Terms—boundary detection, endobronchial ultrasound, DBSCAN, moving average filter, cubic spline interpolation, polar system

Cite: Wannapon Suraworachet, Rajalida Lipikorn, and Anan Wattanathum, "Pulmonary Lesion Boundary Detection from an Endobronchial Ultrasonogram Using Polar Sector Maximum Intensity," International Journal of Electronics and Electrical Engineering, Vol. 4, No. 2, pp. 185-188, April 2016. doi: 10.18178/ijeee.4.2.185-188