Introduction Proper anatomy learning during medicine undergraduate programs is crucial for future clinical correlation. Different techniques provide worthy approaches for this purpose, but exhibit disadvantages such as odors and irritating chemicals in cadavers or spatial perception loss in association with medical images. A complementary technique for spatial‐reasoning learning is proposed using 3D printed models of the aorta artery during an anatomy class. The comprehension of gross characteristics like real dimensions and shape of this main vessel are essential. Therefore, providing a real shaped model could represent a complementary tool for learning and understanding this anatomy. Methods Computed Tomography (CT) scans from VISCERAL challenge were used with their annotations to obtain virtual 3D models of the aorta, this dataset was divided in training and validation. Using Matlab 8.5 an automatic aortic segmentation algorithm was developed through histogram matching to normalize gray intensities within the dataset. A 3D implementation of watersheds segmentation was applied to the CT volumes using markers inside and outside the aorta. Inside markers were selected semi‐automatically using the annotations and automatically, based on the information from the training set. The Dice coefficient was used to compare the real and obtained aortic segmentations. Four aortic segmentations were scaled to real size and exported as stereolitography files for 3D printing. Printed models were used during a second year 62 students' theoretical anatomy class. A semi‐quantitative survey was conducted at the end in which their perceptions related to the model usage, benefits and disadvantages were inquired. The survey was previously validated by experts and had 3 open questions, 2 yes/no questions and 4 one to five scale questions. Results Each aortic segmentation was obtained in less than 8 minutes. Average Dice of 61% was obtained using semiautomatic markers and 40% using automatic markers due to aorta‐related structures considered relevant for the printed models. 80% of the students consider the models as adequate for learning. The students perceived the models provide better structural and spatial understanding of the aorta in comparison with other learning tools (83%). 51% suggested adding related structures (missing data 4,8%). 56% considered the models lack physiology and nomenclature input. 53% assessed the models as a more practical tool compared with imaging learning software. 63% stated optimal anatomical detail and 88% consider the models offer good or excellent usefulness. Conclusions An accurate aortic segmentation algorithm for gross anatomy education was developed. Students showed remarkable acceptance, considering these as useful tools to understand real dimensions and enhance visual perception. Other printed structures should be created to understand systemic correlation and specific anatomical details. Further research is needed in order to understand students' interaction with these spatial‐reasoning learning tools. Support or Funding Information Printing material provided by Universidad de los Andes