To overcome the challenges presented by conventional coal gangue images, which often lack significant textural and color differences and are prone to noise interference (such as dust), making it difficult to extract critical information about their internal structure and elemental distribution through standard imaging methods, a high-resolution CT imaging dataset of coal gangue with strong anti-interference capabilities has been developed. This dataset allows for a more detailed analysis of the internal structure of coal gangue. Furthermore, to address the issue of low classification and recognition accuracy of traditional coal gangue images using deep learning algorithms, a new algorithm for analyzing the elemental composition of coal gangue based on CT images is proposed. The algorithm utilizes an enhanced Res-Unet semantic segmentation model to segment the elemental regions within CT images and analyze their proportions, enabling effective classification and recognition of coal gangue. The model incorporates an efficient local attention (ELA) module within the Res-Unet encoder, allowing it to focus more on important features. Additionally, improvements to the skip connections in the Res-Unet model facilitate better information fusion across different scales, significantly boosting segmentation performance and ensuring accurate delineation of elemental regions in coal gangue CT images. Experimental results demonstrate that the enhanced Res-Unet model successfully segments elemental regions, achieving an mIOU of 84.48%. By calculating the proportions of elemental regions for the final classification of coal gangue CT images, the improved model achieves a classification accuracy of 94.4%, outperforming other models. These results confirm the effectiveness of the proposed algorithm for analyzing the elemental composition of coal gangue based on CT images. This algorithm provides a novel approach and methodology for coal gangue image classification, offering valuable technical support for intelligent coal sorting in factories and promoting the advancement of intelligent and automated systems in the coal industry.