Lung cancer is the deadliest disease among other lung diseases. Appropriate as well as timely treatment of lung cancer is essential to increase survivability of the affected individual. Generally, lung disease is monitored by utilizing CT images. Yet, manual evaluation of lung diseases via CT images is challenging demands additional resources, and requires making appropriate decisions. However, proper analysis of symptoms is significant to make an effective decision based on the treatment strategies of the cancer tissues. Recently, numerous methods with artificial intelligence are frequently used in multiple tasks, including detection, classification, etc., due to their ability to perform early diagnosis with high detection accuracy. However, it is significant to design an automatic diagnosis framework by leveraging the benefits of deep-learning strategy to handle the substantial quality of information to attain higher detection accuracy than traditional frameworks. Therefore, an automated lung cancer detection system is introduced with an effective segmentation process. Here, the required images are fetched from benchmark available resources. These images are then involved with the segmentation, where the Transformer-based Inception U-net++ (Trans-IUnet++) is utilized for segmenting the abnormalities. This model is suitable for identifying the tiny and subtle changes present in the images, thus improving the network's performance. Then, the resultant segmented images are given to the classification stage by Adaptive Dense Inception Net with Spatial Attention (ADINet-SA) is used for classifying lung cancer. In this network, a Modified Uniform Random Number-based Sculptor Optimization Algorithm (MURN-SOA) is included for optimizing the DINet model's hyperparameters, which improves the overall effectiveness. Finally, the designed method's performance is examined and compared with standard techniques to underscore its supremacy.
Kale et al. (Sun,) studied this question.