Asian Journal of Mathematics and Computer Research

This study investigates a novel hybrid technique that integrates deep learning and recurrent neural networks to identify lung cancer in chest X-ray images. The researchers leveraged a dataset of 247 chest X-rays obtained from the Kaggle platform's JSRT (Japanese Society for Radiological Technology) collection. This dataset included 56 X-rays with confirmed lung cancer and 191 without. The inserted image was processed by changing its size to 480 × 480 pixels with the same extension (png) without affecting the image quality. Normalization is an interesting advanced processing step in image processing applications, and data augmentation techniques have also been used to make the data look more diverse. Three incremental strategies have been used to create new training groups, called sphenic conversion (rotation, shear, scale). Then CCN's VGG16 development was used with higher generalization and accuracy compared to other networks in terms of extracting different level features, producing a vector with all the distinctive characteristics associated with LSTM which keeps data in weights form during the training to do the classification. The proposed technique achieved impressive recognition rates, with a testing accuracy of 90.99%. This result translates to highly accurate detection and classification of lung cancer in chest X-rays, differentiating between cancerous and non-cancerous lung images. These findings highlight the promising potential of deep learning for various medical applications. Index Terms Deep Learning, Recurrent Neural Network, Lung Cancer, Convolutional Neural Network, Long Short-Term Memory.