Development of TransUNet-based Segmentation and Residual Attention Network-Derived Lung Cancer Detection

Lung Cancer (LC) is determined as the most common cause of death in recent times all over the world. The only way to decrease the mortality count has been made possible only through earlier detection to save the lives of people. It may affect both men and women. There is much machine learning and Deep learning used to detect cancer in the earlier stage. In this case, image processing has been utilized all over the world, especially in various medical regions for image enhancement in the treatment phase as well as an earlier detection technique. Here, the time factor has played a crucial role in discovering the abnormality problems on the target images. Then, it is particularly used on different cancer tumors like LC. In this paper, the LC detection technique using segmentation is introduced. In the beginning phase, the images related to the cancer detection processes have been congregated from the dataset. Then, the congregated images are fed into the pre-processing phase, which takes a essential part in pre-processing the relevant images to be useful for further phases, where the Contrast Limiting Adaptive Histogram Equalization (CLAHE), histogram equalization, median filtering techniques for attaining the pre-processed images. Further, it has been given to the segmentation phase, where the TransUnet techniques to utilized to segment the images in a proper manner. Then, it has been given to the final phase known as classification is used to classify the patients whether it is normal or abnormal. Further, the classification phase used the Residual Attention Network (RAN) technique. In the end, the recommended LC detection model using segmentation is developed and thus shows better outcomes.