Abstract

One of the malignant tumours with the highest mortality rate and closest to our own mortality is lung cancer. It is extremely dangerous to human health and mostly affects smokers. Lung cancer incidence is rising steadily in our nation as a result of the acceleration of industrialisation, environmental pollution, and population ageing. Computed tomography (CT) pictures are a frequently used visualisation tool in the diagnosis of lung cancer. Using X-ray absorption to create a picture, CT scans can see all types of tissues. Pulmonary nodules are the collective term for the diseased lung tissue; each type of nodule has a unique shape, and each type of nodule has a unique risk of developing cancer. Because the computer vision model can swiftly scan every area of the CT image of the same quality for analysis and is unaffected by tiredness or emotion, computer-aided diagnosis (CAD) is a particularly ideal way to address this issue.

Computer vision models may now assist doctors in diagnosing a variety of ailments thanks to recent advancements in deep learning, and in certain instances, models have even outperformed medical professionals. The use of computer vision in medical imaging detection of diseases has significant scientific significance and value based on the prospect of technical advancement. In this study, we tested the efficacy of a deep learning-based model using CT scans of lung cancer to accurately and promptly detect long illness. The three components of the proposed model are (i) lung nodule detection, (ii) false positive reduction of the discovered nodules to remove “false nodules,” and (iii) categorization of benign and malignant lung nodules. Additionally, several network architectures and loss functions were created and implemented at various times. Additionally, Noudule-Net, a detection network structure that combines U-Net and RPN, is presented to enhance the accuracy of the proposed deep learningbased mode and the identification of lung nodules. The proposed technique has significantly improved the expected accuracy and precision ratio of the disease under consideration, according to experimental observations