Systems, Man, and Cybernetics

 

Home

• [P768] Pulmonary Nodule Classification Aided by Clustering
  Author(s): S. L. A. Lee, Abbas Z. Kouzani, Gulisong Nasierding, E. J. Hu
  
  

  Abstract
  Nodule detection in lung images is an important component of lung cancer detection systems. Ensemble based learners have the potential to perform better than non-ensemble based learners. Random forest (RF) based classifier is employed to pulmonary nodule classification where each tree produces a classification decision, and an integrated output is calculated. In addition, a classification aided by clustering (CAC) approach is implemented to improve the classification performance. Three experiments are devised using LIDC database of 32 cases with 5721 slices. The classification performance and execution times are presented and discussed. Overall, the best sensitivity of 97.08% and specificity of 95.45% is recorded for RF based CAC which outperforms SVM and DT based CAC.

  
   

Full Text (PDF)