Performance Analysis of Modified CNN Model In Lung Nodule Segmentation

Cancer is a major public health concern on a worldwide scale. It was demonstrated that lung cancer has the greatest rates of morbidity and mortality in both males as well as females. Segmenting the lung nodule is crucial for lung cancer treatment. This work's purpose is to segment such lung nodules utilizing CT (Computed Tomography) scans. Biological image segmentation is achieved by combining the UNet with a modified convolutional neural network. To segment lung nodules, this study employs customized versions of CNN. To obtain the ideal result, this study also employ a range of optimizers. The dataset was acquired from LIDC-IDRI, and these CT scans were pre-processed to produce images of nodules. The proposed model helps to reduce the run time of the automated lung cancer nodule segmentation system. To acquire nodule images and the masks, these CT scans undergo additional preprocessing. The architecture is used to train the pre-processed images, and the performance of the model is computed. For the suggested CAD (Computer-Aided detection) system's segmentation performance, a modified CNN model yielded a DSC (Dice Similarity Coefficient) value of roughly 0.8205 using the Adam optimizer. This experiment yields better lung nodule segmentation results as compared to other recent approaches.