Early detection of lung nodules is important since it increases the probability of survival for the lung cancer’s patient. Conventionally, the radiologists will manually examine the lung Computed Tomography (CT) scan images and determine the possibility of having malignant nodules (cancerous). This process consumes a lot of time since they have to examine each of the CT images and marking the lesion (nodules) manually. In addition, the radiologist may experience fatigue due to large number of images to be analysed. Therefore, automated detection is proposed to assist the radiologist in detecting the nodules. In this paper, the main novelty is the implementation of image processing methods to segment and classify the lung nodules. In this work, several image processing methods are utilized namely the median filter, histogram adjustment, Inverse Surface Adaptive Thresholding (ISAT) to segment the nodules in CT scan images. Then, 13 features are extracted and given as input to the Back Propagation Neural Network (BPNN) to classify the image either benign or malignant. Based on the result obtained, it showed that ISAT segmentation achieved 99.9% in term of accuracy. The extracted features were given as input to the Back Propagation Neural Network (BPNN) to classify the image either benign or malignant. Lung nodules that are less than 3 mm are considered as benign (non-cancerous) and if their size is more than 3 mm, there are considered as malignant (cancerous). The results showed that the proposed methods obtained 90.30% in term of accuracy.
