3DCGRU: THREE-DIMENSIONAL CONVOLUTIONAL NEURAL NETWORK-BASED BIDIRECTIONAL GATED RECURRENT UNIT MODEL FOR LUNG CANCER CLASSIFICATION

In the current decade, the mortality rate is expanding as a consequence of lung cancers, which occur due to abnormalities and inappropriate development of the lung tissues that led to lung cancer all over the world. Prior lung cancer classification and detection diminishes the death rate of lung cancer. Various deep learning methods and traditional methods were implemented for detecting and classifying lung cancers, but such methods had challenges in implementing multiple models and failed to detect the different types of cancer nodules and smaller tumors, time complexities, minimal datasets, and reduced robustness. Thus, to overcome such limitations, the research developed a three-dimensional convolutional neural network-based gated recurrent unit (3DCGRU) model for the effective classification of lung cancers. For tuning the parameters, the method used the Vulpes Hunt Optimization (VuHO) algorithm with the ability to observe and exploit. Through the incorporation of [Formula: see text]-fold and training percentage, the performance is analyzed. The superior performance of the proposed model is examined using the Lung-PET-CT-Dx and LUNA 16 dataset, where the proposed model reveals an enhanced performance with an [Formula: see text]1-score of 95.36%, accuracy of 95.68%, as well as recall of 93.77% and precision of 93.77% for 80% of training. Moreover, the model reveals an [Formula: see text]1-score of 96.13%, an accuracy of 95.74%, a recall of 94.14% and a precision of 94.14% for [Formula: see text]-fold-10 with the Lung-PET-CT-Dx dataset.