Detection of Cancerous Nodule in Lung Using KNN Classifier

Author Name(s): Sakshi Wasnik, Pallavi Parlewar, Prashant Nimbalkar
Author Email: wasnikss_1@rknec.edu

Abstract

The paper focuses on detection of cancerous nodule in lung using KNN classifier for the medical images to help the physicians to interpret the computer tomography (CT) scan images. The combination of various image processing tools to effectively accomplish the required objectives. Pre-processing of an image is a done to remove the noise in the image. Threshold value is calculate using Otsu thresholding. Tumor is extracted by thresholding the image and is subtracted by the removal of small object and is extracted from the fore ground. The features like area, perimeter, major axis length, minor axis length, eccentricity and convex area are obtained. K nearest neighbour algorithm is used to find the cancerous tumour. The proposed algorithm uses —- samples for each iteration and achieved accuracy 96.25%, sensitivity 98.36% and specificity and 89.47%.

Introduction

Lung cancer is the most common cancer across the world. To save human lives it is very necessary to detect the lung cancer at an early stage. CT is one of the modest medical imaging methods to diagnose the cancer. The lungs are the essential part of the body that is use to breathe [1]. Lung cancer, also known as lung carcinoma, is a malignant tumour characterized by uncontrolled growth of the cell. It is therefore necessary to treat this to avoid spreading its growth to other parts of the body [2]. Long-period tobacco smoking is the primary factor for 85% of lung cancers [3]. Computer tomography (CT) are the conventional methods to detect the presence of lung cancer. Hence, it is essential to determine a new robust method to diagnose the lung cancer at an earlier stage [4]. The system is separated into four stages. In the first stage enhancing of CT images is done by the usage of enhancement techniques like median filter. Thus the noise in an image gets reduced and also smoothens the edges. To separate out various parts from the image, so that the tumor is clearly visible, segmentation is done. Then various features of an image are computed such as area, perimeter, major axis length, minor axis length, convex area and eccentricity so that it can be used for training of the system. After that KNN classification algorithm is applied to detect the cancerous tumour. To surpass the system, there is a need to experimentally test all the tumour and non-tumour images from the dataset.

Conclusion

This work discusses the need for techniques and tools that help doctors optimize the screening and diagnostic workflow of lung cancer to improve the precision of cancer diagnosis and decrease exposure to costs and radiation. The proposed algorithm is tested under various condition and found to be 96.25% accurate for classifying the nodule using KNN classifier. The sensitivity and specificity of the proposed system is 98.36% and 89.47% respectively

763 total views, no views today

Download PDF File

About the author: admin