Robust cell segmentation for non-small cell lung cancer

Lung cancer, predominantly non-small cell lung cancer (NSCLC), is one of the most serious and life threatening cancers, and automated, accurate cell segmentation for NSCLC is a prerequisite for many subsequent quantitative analysis on digitized images. However, the complex nature of histopathological images presents significant challenges for most traditional segmentation algorithms. In this work, we propose a novel and robust touching cell segmentation algorithm for NSCLC, which first localizes cell seeds using distance transform-based voting and thereafter employs a repulsive balloon snake model to accurately segment cells with initialization using the detected seeds. The comparative experiments with the recent state-of-the-art demonstrate the effectiveness of our method.