ICNN-Bi-LSTM framework for lung cancer classification in CT imaging with statistical and texture pattern extractors

Abstract One of the deadliest and most terrible diseases in the world is lung cancer; however, prompt analysis and care may safeguard lives. When a person with lung cancer does not exhibit early symptoms, it is very challenging to determine and predict the affected area. Despite being the most effective imaging method in the medical sector, CT scan images are more challenging for physicians to interpret and diagnose diseases. Consequently, computer-aided diagnostics can be useful in helping medical professionals to precisely detect cancer cells. This research introduces a new lung cancer classification (LCC) model. The initial pre-processing is performed with Adaptive Weiner Filtering (AWF), followed by Modified Deep joint (DJ) Segmentation, which is processed to segment the ROI and non-ROI regions. Then, Local Gradient Increasing Pattern (LGIP), Median Binary Pattern (MBP), Multi-texton (MT) features, and Statistical features are extracted. Then, classification is done using HC that involves an Improved Convolutional Neural Network (Imp CNN) and Bi-LSTM models. Finally, investigation outcomes on wide-ranging metrics show the performance of the proposed model. The proposed LCC model achieves higher classification accuracy compared to existing state-of-the-art lung cancer classification approaches, confirming its superiority and effectiveness.