Document Type: Research Articles
Department of Electronics and Communication Engineering, Bannari Amman Institute of Technology, India.
Objective: Lung cancer is a type of malignancy that occurs most commonly among men and the third most common
type of malignancy among women. The timely recognition of lung cancer is necessary for decreasing the effect of
death rate worldwide. Since the symptoms of lung cancer are identified only at an advanced stage, it is essential to
predict the disease at its earlier stage using any medical imaging techniques. This work aims to propose a classification
methodology for lung cancer automatically at the initial stage. Methods: The work adopts computed tomography (CT)
imaging modality of lungs for the examination and probabilistic neural network (PNN) for the classification task.
After pre-processing of the input lung images, feature extraction for the work is carried out based on the Gray-Level
Co-Occurrence Matrix (GLCM) and chaotic crow search algorithm (CCSA) based feature selection is proposed.
Results: Specificity, Sensitivity, Positive and Negative Predictive Values, Accuracy are the computation metrics used.
The results indicate that the CCSA based feature selection effectively provides an accuracy of 90%. Conclusion: The
strategy for the selection of appropriate extracted features is employed to improve the efficiency of classification and
the work shows that the PNN with CCSA based feature selection gives an improved classification than without using
CCSA for feature selection.