Early Lung Cancer Detection Using Artificial Neural Network

T. Pandiangan, I. Bali, A. R.J. Silalahi


Lung carcinoma is a malignant lung tumor that is deadly and is characterized by the uncontrolled cell growth in the tissue of lung. Normally the lung cancer detection is done by visual inspection of x-ray image by medical doctor. The purpose of this study is to create a computational tool that can automatically detect early lung cancer from x-ray image. This research has two main steps, with first being to characterize cancer or cancer symptoms based on x-ray images and second step is to develop an artificial neural network (ANN). In first step, particularly it is wanted to lay out a rigorous image processing framework with sequential steps: (i) image noise reduction, (ii) image enhancement, (iii) lung organ segmentation, (iv) object edge detection, and (v) tumor boundary detection. The framework incorporates image processing techniques such as thresholding and morphological detections (erosion and dilation). The framework is expected to reveal the relevant features that define lung cancer or early lung cancer such as area, perimeter, density profile and shape ratio. For the second step, the ANN is built based on machine learning algorithm to study a large set of x-ray images of positively diagnosed lung cancer patients. In addition to learning solely based on the 2D x-ray images, it is also incorporated the previously studied tumor features. The two combined with a large dataset is expected to enable the machine to reach a close to 100 % detection accuracy. Based on the test results of 10 samples obtained the comparative value of the calculated by the ANN with the results of measurement with Matlab program is tends to approach the same. It can be concluded that ANN has been successfully educated so that can identify 10 samples correctly.


Malignant tumor, Image processing, Enhancement, Neural network, X-ray

Full Text:


DOI: https://doi.org/10.17146/aij.2019.860

Copyright (c) 2019 Atom Indonesia

Creative Commons License
This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.