Lung cancer causes the higher number of deaths each year of any type of cancer. Generally this cancer is detected when no possible to perform an effective treatment and the usage of computational methods for segmentation can improve the time to initiate this treatments. The medical segmentation decathlon is a challenge to do a generic segmentation method for 11 different tasks. However, the current methods have poor results on the lung nodule segmentation task. This work focuses on implementing changes made by state of the art methods in lung nodule segmentation to the network developed by the Biomedical Computer Vision group from the Universidad de los Andes for the challenge. All of this reducing the computational cost of the network. The final network implemented, a deeper model of the original referred to as ROG+ reduce significantly the computational cost of the method while maintaining similar but lower metrics with a 27.58% dice score.