DeepFusion-LC: A Multi-Model Explainable Framework for CT-Based Lung Cancer Classification

Abstract

Due to the increased rate of lung cancer patients and the urgent need for early and accurate detection, intelligent and automated medical image processing is in great demand. Conventional techniques for medical image processing heavily depend on the expertise of medical professionals for proper analysis. In addition, such conventional techniques are likely to cause delays in proper detection and may not be effective for a large number of patients. Keeping such factors in view, a sophisticated framework for classification of lung cancer is presented in this paper. The framework is based on effective implementation of deep learning techniques to improve the accuracy and reliability of the classification process. In the proposed framework, computed tomography images are subjected to preprocessing to improve image quality and remove noise. The images are further processed using a sophisticated feature extraction technique based on convolutional neural networks. The technique is effective in handling different complexities in image processing, such as data imbalance, tumor size, and imaging conditions. The features are further subjected to classification to effectively classify lung conditions into different classes. In addition, the framework is based on effective evaluation to improve accuracy and reliability. The proposed framework is effective in comparison to conventional and existing techniques based on deep learning for image processing. The framework is developed to help in early detection of lung cancer and assist medical professionals in decision-making.