Rodrigo Bastidas-ChalánDiego Hernández-YánezMoisés Mora-MurilloPaúl Valencia-OrdoñezFélix Vargas-Hincapié2026-03-202026-03-2020269783031987670https://doi.org/10.1007/978-3-031-98768-7_15This research develops a classification model for lung tumors based on the fractal dimensions of their contours, based on the premise that the surface texture of carcinomas, characterized by their fractal dimension, can be correlated with the clinical behavior of the tumor. A systematic review of the literature on lung carcinomas and the use of artificial intelligence techniques in medicine was carried out. Using DICOM medical images from the National Cancer Institute and clinical data in CSV format, the images were converted to NIFTI and processed in Python to obtain fractal dimension measurements, which were integrated with clinical data in a database for statistical analysis. Three fractal measurements were calculated for each tumor: DF1, DF2 and DF3 with values between 1.4 and 2.0, suggesting that tumors with more complex contours could be more aggressive and in advanced stages of the disease. By testing various machine learning algorithms, the most effective neural network achieved 56% accuracy in histology classification after 5000 training epochs. Although no significant correlations were found with age and gender, it was observed that tumors with higher fractal dimensions were associated with more advanced stages and shorter survival times. These results suggest that the fractal dimension may be valuable for the classification and prognosis of lung tumors. However, more research is required to improve the accuracy of the models and better understand the observed correlations.enDICOMFractal dimensionRandom forestTomographybook-chapter