Development and implementation of a vision transformer based approach for lung cancer diagnosis from computed tomography images

Lung cancer is a fatal disease in which early diagnosis is of vital importance. Manual evaluation of computed tomography (CT) scans is often challenged by human errors and inter-observer variability, which complicates the diagnostic process. Therefore, the integration of computer-aided diagnosis systems into clinical workflows has become a significant necessity. I n this thesis, vision transformer architectures are examined for the classification of lung cancer from CT scans, and a new deep learning architecture named FocalNeXt is presented. FocalNeXt combines the attention mechanism of FocalNet with the feature extraction capability of ConvNeXt, forming a powerful Vision Transformer-based structure. The model was tested on the IQ-OTH/NCCD dataset and achieved an accuracy of 99.81%. It also demonstrated superior performance in terms of precision, recall, and F1 score. The results indicate that FocalNeXt offers an effective and reliable solution for lung cancer detection and outperforms existing methods in classification performance.