OLDC-NET: OPTIMIZED RECURRENT CONVOLUTIONAL NEURAL NETWORK-BASED LUNG DISEASE DETECTION AND CLASSIFICATION
Lung cancer is a critical illness that kills millions of people worldwide. As a result, early diagnosis and categorization of lung tumors has the potential to save millions of lives. However, traditional approaches failed to provide superior categorization results. Thus, this article is focused artificial intelligence approach with optimized lung disease classification network (OLDC-Net) for multi class classification. Initially, hybrid recursive box filtering (HRBF) is used to perform the preprocessing of the Computed Tomography (CT) based lung images. Then, segmentation of lung cancer is performed using Unified-K-Means clustering (UKMC) operation, which locates the cancer effected region. Further, features are extracted using multi-level discrete wavelet transform (ML-DWT), which contains the disease specific information. Finally, natural inspired moth-swarm optimization algorithm (MSOA) is used for feature selection operation, which select the best features from available features. Finally, recurrent convolutional neural network (RCNN) is used to perform the classification of lung cancers with benign, malignant lung types. The simulation results shows that the proposed OLDC-Net resulted in superior segmentation, classification performance as compared to conventional methods.
Computed tomography lung images, lung disease detection and classification, hybrid recursive box filtering, K-Means clustering, K-Means clustering, recurrent convolutional neural network.