Utilization of U-Net Deep Learning for Automated Lung Tumor Detection in 3D CT Scan Image Segmentation

Abstract

Lung tumors are a serious health problem with a high mortality rate, requiring accurate and efficient detection methods. Thoracic CT scans are the primary imaging modality for lung tumor evaluation due to their ability to produce three-dimensional images. However, manual tumor segmentation is time-consuming and subject to observer variability. This study aims to utilize deep learning with a U-Net architecture for automated lung tumor detection in 3D CT scan image segmentation. This study employed a quantitative experimental approach. The data consisted of thoracic CT scan images accompanied by manual segmentation masks as ground truth. The research stages included preprocessing, training of a 3D U-Net model, and evaluation of segmentation performance using the Dice coefficient. The results of the study show that the 3D U-Net model is capable of automatically segmenting lung tumors with a good level of agreement with manual segmentation. The model achieved an average accuracy of 0.99, indicating its ability to accurately identify the lung tumor area. In addition, the model is also able to consistently identify the size of the lung tumor and enhance the efficiency of the image analysis process. In conclusion, U-Net-based deep learning is effective for automated lung tumor segmentation in 3D CT scan images and has the potential to support radiologists in improving diagnostic accuracy and efficiency.