Deepa P L

M.tech

Mar Baselios College of Engineering and Technology

Asst. Professor

NLST (Learn more about this study)

NLST-632

Jan 27, 2020

Lung cancer detection using Deep Convolutional Neural Network

The increased use of technology has made an overwhelming impact on the overall well-being of a person. Health experts are increasingly taking advantage of the benefits of these technologies, thus generating a scalable improvement in the area of health care. Because of this, there is a paradigm shift from manual monitoring towards more accurate virtual monitoring with the minimum percentage of error in the area of health care. Advances in artificial intelligence (AI) have led to exciting solutions with high accuracy for medical imaging technology and is a key method for enhancing future applications in health care. Lung tumor detection is an important task in medical image processing. Early detection of lung tumors plays a crucial role in improving the treatment possibilities and increases the survival rate of the patients. Manual detection of the lung tumors for cancer diagnosis, from a large amount of CT scans generated in clinical routine, is a difficult and time-consuming task due to the complexity and the vast amounts of knowledge and skill needed in a particular subspecialty of radiology. So, there is a need for automatic lung tumor detection from Lung CT-scan images. Deep learning methods can achieve this task. Different deep learning networks can be used for the detection of lung tumors. Here we are planning to create a new Deep Convolutional Neural Network for lung cancer detection and classification. We are trying to detect the cancerous area from the CT scan images. Also, we need to classify it into different types of lung cancers. This network can effectively identify the presence of the tumor area and segment it out.

Here we need an enormous amount of lung cancer data, both normal and having tumors for doing the detection problem. Also, we need data for classification networks too. Tumorous data are freely available, but normal lung cancer images are not available. At least 1000000 images are required for the training purpose for increasing the effectiveness of the created network. As the number of data increases, detection and classification accuracy will increase.