ARG-Based Malignancy Evaluation for Pulmonary Nodule Detection

Malignancy evaluation is an important stage in lung cancer diagnosis. However, due to the complexity of medical image data, processing the images costs a long time. Aimed for this problem, we propose a method based on the Attributed Relational Graph (ARG) in this paper to measure the degree of pulmonary nodule malignancy, with And-Or Graph (AOG) and Markov Blanket (MB) used for dimensionality reduction. Our method is tested on real images of Shanghai Chest Hospital and Shenyang Shengjing Hospital with classification accuracy 70.85%, and tested on the IDRI-LIDC database with accuracy 75.01%. The method can help to identify and analyze the benign and malignant nodules correctly and quickly.