30 CT scans from 30 different pediatric patients diagnosed with any type of lymphoma were collected. All patients were hospitalized between the years 2013 and 2020. One CT scan per patient was included to maximize potential variance in the data and increase model generalizability. There were 17 boys and 13 girls in the study group, and the mean age was 12.8 (SD 4.1, minimum age 2, maximum age 17). Subsequently, two radiology residents performed segmentation of thoracic lymph nodes in 3D Slicer software. The first 17 segmentations were performed manually. Then, an initial model was trained on these scans and used to perform initial segmentations on the remaining 13 scans to accelerate the process. All initial segmentations were subsequently reviewed by the radiology residents responsible for segmentation, and necessary adjustments were made. Dataset consists of raw CT scans and accompanying segmentation. Dataset publication and collection were funded by Polish National Science Centre grant 2020/36/T/NZ5/00515