Abstract

There is a consensus that accurate and timely diagnosis of lung cancer is crucial, and it is considered a serious condition, essential for effective treatment of affected individuals. Conversely, several previous studies have examined issues related to automated AI-based systems involving diagnostic results, aiming to diminish the difficulties related to human observation by integrating computer-aided diagnosis (CAD) systems into the diagnostic workflow to enhance efficiency and improve outcomes. This paper discusses the impact of image processing techniques on automated lung cancer detection using computed tomography (CT) scan images and demonstrates their key role in improving model performance. A pre-trained U-Net model was used in conjunction with a set of Gaussian difference (DoG) filters and a CLAHE filter. A detailed examination of a publicly available CT dataset from the Iraqi Teaching Oncology Hospital/National Cancer Center (IQ-OTH/NCCD) included careful comparisons regarding the training efficiency between image processing and non-processing methods. The results achieved were of high quality and satisfactory with the proposed methodology. The results with processing achieved an accuracy of 99.8%, a precision of 99.5%, and F1 score of 99.8%, while without processing, it achieved an accuracy of 97.1%, a precision of 95.2%, and F1 score of 96.11%. These results demonstrate the potential to advance personalized medical imaging and life-saving care.