Cancer is still the leading cause of death in the world today. Lung Cancer in particular, the second most common type in Brazil, has its own visual characteristics in imaging exams. The main non-invasive exam to detect pulmonary nodules is computed tomography in the chest region. In this three-dimensional exam, experts analyze the results in “slices” available in 3 axes (axial, coronal and sagittal), where each axis is made up of hundreds of two-dimensional slices. Considering the volume and complexity of the data analyzed, one way to further improve and speed up the analysis results is to make use of artificial intelligence tools, which can support the diagnostic decision-making process. In addition to the potential for clinical use of the tool, there is also the possibility of use for research. The present work aims to study and evaluate tools to automate the process of detecting lung nodules in three-dimensional exams in a reproducible way. To develop the research, different public databases were used, including one from the Center of Image Sciences and Medical Physics of the Hospital das Clínicas of the Ribeirão Preto Medical School, which contain both the anonymized exams and the corresponding annotation of the nodules. A single-step model was implemented and evaluated to perform volumetric object detection for different exam Hounsfield Unit (HU) windowing. The generated models are used together as a committee of classifiers to return the desired final result (extremities coordinates of the found nodes). The results obtained with the test data set (sensitivity of 91.69% with an average rate of 1.00 false positives per exam) indicate great proximity to works with the same purpose in the literature, and, therefore, we can say that the developed algorithm has great potential for use, both in clinical and academic settings.