MedNet: A Segmentation Algorithm for Effective Lung cancer Diagnosis

The detection of lung cancer is difficult and also impeded by the use of computed tomography for humans. Due to human error rate, the prediction is inaccurate. With the advanced techniques such as Deep learning, the presence of cancer can be determined and segmented effectively. Improvised Deep learning models in artificial intelligence have generated incredible outputs as well as the traditional methods in various fields in the past analysis. Presently, many researchers are interested in developing various deep learning methodologies to accumulate and elevate the performance of various systems in detecting lung diseases along with computed tomography images. But the existing algorithms show lesser accuracy in performing segmentation of the cancerous part. In this proposed work, a novel algorithm for separating the cancer affected area in the lungs. The segregation of cancer infected area from unaffected part of lungs in computed tomography image is known as segmentation. In this proposed system, a segmentation algorithm called MedNet is developed. In this research work, MedNet is introduced and trained on datasets of lungs that contains computed tomography images of lungs to execute segmentation of lung cancer. Three different layers such as 30 convolutional layers, 3 bidirectional convolutional layers and 3 maxpooling layers are involved in MedNet and proposed in such a way to overcome the existing disadvantages effectively and to increase the accuracy.