Auto segmentation, detection, and countering of Lung nodule using convolutional neural networks

Lung cancer is one of the most common types of cancer and leading causes of cancer-related deaths in the United States. Whereas radiotherapy is often applied to treat lung cancer, delineation of the tumor area known as planning target volume (PTV) is a time-consuming and critical task. In this paper, we have proposed a method based on convolutional neural network for an auto detection of PTV to aid radiation treatment for lung cancer. Using a gradient descent search algorithm and convolutional neural network for classification, we can detect lung tumors and nodules. In doing so, pre-processing and segmentation are performed on lung CT images for obtaining the segmented tumor and non-tumor region. The trained algorithm results will be compared to pre-segmented and contoured CT images for validation and accuracy. The accuracy of our algorithm is 64 % in detecting lung nodules. We believe that the use of such algorithms will help in improving the clinical flow and efficiency of radiotherapy for lung cancer while reducing the chances of a missed diagnosis resulting in increase in the life expectancy of the patients.