Lung Cancer Prediction using Recurrent Residual Convolutional Neural Network and Long Short Term Memory

Abstract

Among Cancers lung cancer is one of the most prevailing and death-dealing cancers and the early detection of lung cancer poses a significant challenge due to its often asymptomatic nature, leading to high mortality rates. We introduce a novel method for precise lung cancer detection from CT images, leveraging deep learning techniques. Our approach combines Recurrent Residual Convolutional Neural Networks (RRCNN) like RU-Net and R2U-Net with Long Short-Term Memory (LSTM) networks and VGG16 architecture. Incorporating LSTM and Recurrent Residual Convolutional layers enhances the model's ability to capture temporal dependencies and improve feature representation for segmentation tasks. Additionally, VGG16's strong feature extraction capabilities aid in identifying complex patterns and subtle abnormalities in lung images. Evaluation on the LUNA16 dataset, comprising 888 CT scans with 1186 annotated lung nodules, demonstrates our method's superiority, achieving an impressive 90% accuracy. Our findings underscore the crucial role of temporal information in distinguishing between benign and aggressive lung cancer cases. In conclusion, our study highlights the effectiveness of combining RRCNN, LSTM, and VGG16 for accurate lung cancer detection, offering promising prospects for early diagnosis and improved patient outcome.