Medical image segmentation has a fundamental importance for diagnosis, treatment planning, and anatomical study in computed tomography (CT). In image segmentation fields, methodologies are broadly categorized into supervised and unsupervised approaches. Supervised segmentation relies on labeled datasets to train models; however, it is heavily dependent on large, high-quality labeled datasets from an expert. Unsupervised methods do not require prior labeling and instead group pixels based on intrinsic image features such as intensity, texture, or spatial proximity, we use the term structural segmentation to refer to this. Within unsupervised segmentation methods, two major subcategories could be defined: region-based and intensity-based approaches. For this work, intensity-based methods were used to classify pixels primarily based on grayscale values, offering a simpler yet effective solution, particularly for CT images. This work presents a comparative evaluation of three different intensity-based segmentation algorithms: Otsu’s thresholding, K-Means clustering, and the Watershed transform. Performance was assessed using ground truth masks from a publicly available database, with metrics such as Jaccard similarity index and Hausdorff distance. The results offer insights into the relative strengths and limitations of each method, guiding the selection of appropriate segmentation strategies in clinical and research settings involving thoracic CT analysis. The purpose of structural segmentation in this context extends beyond visual interpretation; it serves as a critical preprocessing step for downstream applications such as region-specific image reconstruction, texture analysis, and feature extraction.