Abstract: Cancer is one of the most serious illnesses in the world, among which lung cancer is the deadliest form. The survival rate of lung cancer in five years is only 54% and the early diagnosis rate is merely 15%. Lung cancer spreads to the different parts of the body rapidly before it is being diagnosed. Therefore the early detection has a crucial role in increasing the survival rate. This paper proposes a vector quantization (VQ) based approach for the detection of lung nodules from computed tomography (CT) scan images. Lungs are extracted from the surrounding tissues using simple thresholding. VQ is performed in two levels, first level of VQ is for lung segmentation and the second level is for the segmentation of lung nodules. Morphological closing operation is performed to refine the lung mask and to ensure the detection of juxta pleural nodules. False positives are reduced using Support vector machine (SVM) classifier. Experimental results shows improved performance comparing to existing computer aided detection (CAD) systems.
Keywords: Vector quantization (VQ), Computed tomography (CT) scan, False positives (FP), Computer aided detection (CAD), Support vector machine (SVM).