LUND-PROBE - LUND Prostate Radiotherapy Open Benchmarking and Evaluation dataset

Abstract

Radiotherapy treatment for prostate cancer relies on computed tomography (CT) and/or magnetic resonance imaging (MRI) for segmentation of target volumes and organs at risk (OARs). Manual segmentation of these volumes is regarded as the gold standard for ground truth in machine learning applications, but to acquire such data is tedious and time-consuming. A publicly available clinical dataset is presented, comprising MRI- and synthetic CT (sCT) images, target and OARs segmentations, and radiotherapy dose distributions for 432 prostate cancer patients treated with MRI-guided radiotherapy. An extended dataset with 35 patients is also included, with the addition of deep learning (DL)-generated segmentations, DL segmentation uncertainty maps, and DL segmentations manually adjusted by four radiation oncologists. The publication of these resources aims to aid research in automated radiotherapy treatment planning, segmentation, inter-observer analyses, and DL model uncertainty investigation. The dataset is hosted on the AIDA Data Hub and offers a free-to-use resource for the scientific community, valuable for the advancement of medical imaging and prostate cancer radiotherapy research.

Fig. 1

Data overview. A hierarchical overview of the dataset and its content for base part (n = 432, blue color, left) and extended part (n = 35, gold color, right). The content in the base part was included in the extended part for each patient.

Fig. 2

Image data types in the dataset. (a) MRI T2w volume with clinical radiotherapy segmentations overlaid. The center of the fiducial marker delineation in black shows the defined center-of-mass for one of the fiducial markers inserted into the prostate. (b) Synthetic CT (sCT) volume with radiotherapy segmentations overlaid. The sCT was created from the MRI T2w images in (a) using the Spectronic MriPlanner software. The high-density sphere in the prostate, seen as a high intensity object, is placed at the fiducial marker center-of-mass point defined in the MRI T2w volume. (c) Synthetic CT (sCT) with radiotherapy segmentations and overlaid dose distribution, see dose color wash legend to the right where 100% equals 42.7 Gy. It is the same patient and corresponding image slice in (a), (b), and (c).

Fig. 3

Prostate segmentations. (a) MRI T2w image slice with prostate CTV deep learning segmentation uncertainty map overlaid. (b) T2w MRI zoomed in with prostate deep learning CTV segmentation. (c) Four different oncologist’s individual prostate CTV segmentations on zoomed in T2w MRI, visualized in separate colors. (d) deep learning prostate CTV segmentation uncertainty map visualized in color, zoomed in from (a). Corresponding data for rectum is available in the dataset.

Fig. 4

Pelvis segmentation overview. One patient volume oriented as demonstrated by the green model. Overlaid are the available prostate targets where PTVT_427 (dark blue) encompasses the prostate CTVT_427 (purple) together with the Bladder (yellow), left (FemoralHead_L) and right (FemoralHead_R) femoral heads (green-yellow), PenileBulb (light blue), Rectum (brown) and Genitalia (red), all available with respective name in the dataset. The BODY segmentation, encompassing the whole scanned patient volume, is the largest volume. Fiducial marker delineations are not shown.