<p>Constipation presents distinct types that require tailored treatments. However, accurate identification of these types remains challenging, complicating intervention selection. The quantity and location of intestinal gas and stool are linked to constipation types, making their quantification clinically significant. Manual quantification of the Gas Volume Score (GVS) is prone to observer variability and imposes a high workload on physicians, while no established method currently exists for quantitatively assessing stool volume. This study proposes an automated method for segmenting gas and stool regions from abdominal X-ray images using U-Net-based architectures. To improve segmentation performance, we evaluate U-Net with backbones such as VGG, ResNet, SE-ResNet, EfficientNet, and employ an ensemble approach. A novel metric, the Stool Volume Score (SVS), is introduced for quantitative stool volume assessment. The study used two datasets with a total of 166 subjects and applied 5-fold cross-validation to ensure robust performance evaluation. The experimental results demonstrated DICE coefficients of 0.767 for gas regions and 0.696 for stool regions. The correlation coefficients between the automated GVS and SVS and manual evaluations were 0.947 and 0.840, respectively. The use of ensemble modeling improved DICE coefficients by 0.002 points for gas regions and 0.012 points for stool regions compared to single models, demonstrating enhanced segmentation performance. This study proposes the first automated method for segmenting gas and stool regions in abdominal X-ray images, offering a robust tool for clinical decision-making in constipation management.</p>