EfficientNet-based deep learning framework for multiclass lung cancer classification using CT scan images

Lung cancer is extremely challenging to treat and detect. It is responsible for more deaths compared to other types of cancer globally. Due to the combination of complex diagnostic techniques and a lack of early detection, many patients only discover they have lung cancer in its late stages. This leads to even more difficulty and complexity in treatment. One of the most effective ways to treat lung cancer is to detect it early. When it first develops, lung cancer is very difficult to detect even with the use of specialized medical imaging. CT scans are the most effective and widely used imaging techniques for lung diagnostics. However, special trained personnel are required to analyze the results. This form of diagnosis is slow and is known to have high variability based on who the specialist is. Deep and machine learning research has made significant progress in the area of automated diagnosis imaging. This paper describes a deep learning approach to automating lung cancer diagnosis using CT imaging. The proposed method is based on EfficientNet and its ability to classify imaging data quickly and accurately. The method is shown to capture the nuances of lung imaging and provide a front-line service for diagnosis to support the clinical staff. The method is also scalable for multiple deployments.