High-resolution computed tomography (HRCT) provides valuable information for the analysis of COVID-19 patients. For the estimation of the damage caused by this disease, most efforts used approaches based on convolutional neural networks (CNN), that require a large amount of labeled data. In this work, a semi-automatic methodology for the volumetric segmentation of pulmonary parenchyma and COVID-19 related lesions based on Probabilistic Active Contours (PACO) is presented, using only intensity information, eliminating the need for large training datasets and expert-labeled data. The proposed method was evaluated using the publicly available COVID-19 Lung and Infection Segmentation Dataset (LISD), achieving a high Dice Similarity Coefficient (DSC) of 0.95 for the segmentation of the pulmonary parenchyma, comparable with CNN-based methods in the state of the art. However, the segmentation of COVID-19 lesions yielded a lower average DSC of 0.26, attributed to challenges in distinguishing small and poorly defined regions. Despite these limitations, PACO demonstrated its potential as a reliable alternative to AI-based methods, offering robustness and repeatability in medical imaging scenarios where labeled data is scarce. Future work will focus on improving lesion segmentation accuracy and exploring the application of this approach in diverse clinical contexts.