Analysis of Underfitting and Overfitting in U-Net Semantic Segmentation for Lung Nodule Identification from X-ray Radiographs

Identifying lung cancer at an early stage can significantly lower mortality rates. To achieve this, segmenting affected areas from lung X-rays is a crucial task. However, developing a network for biomedical image segmentation poses a major challenge due to the lack of labeled data. In this paper, we present a semantic segmentation network using a modified u-net and prepare labels and masks from raw X-ray images for experimentation. Our approach generates a segmented lung nodule or mass picture, and we perform repeated experiments to achieve optimal performance parameters. To further analyze the proposed module, we conduct additional experiments to assess underfitting and overfitting by varying the epoch. The results, based on the Intersection of Union score (IOU) and segmented image clarity, demonstrate the efficiency of our model.