Description
SegRap2025 Dataset will consist of CT images collected by Siemens CT scanners with the following scanning conditions: bulb voltage, 120 kV; current, 300 mA; scan thickness, 3.0 mm; resolution, 1024 × 1024 or 512 × 512; injected contrast agent, iohexol (volume, 60~80 mL; rate, 2 mL/s; without delay). The dataset consists of clinically required non-contrast CT images (ncCT) and contrast CT images (ceCT) from patients with nasopharyngeal cancer before treatment.
The dataset consists of clinically required non-contrast CT images (ncCT) and contrast CT images (ceCT) from patients with nasopharyngeal cancer before treatment.
- Training data will consist of CT images from 120 patients with a corresponding label map, as well as 500 unlabeled cases.
- Validation data will consist of CT images from 20 patients.
- Testing data will consist of CT images from two cohorts: 60 patients from internal testing cohort, and 60 patients from external testing cohort.
Note: All GTVs were annotated individually by oncologists using MIM Software and ITKSNAP, the annotation of each organ was also stored individually. The expected output from your algorithm should be a set of label maps.
SegRap2025 Dataset will consists of CT images from Sichuan Cancer Hospital are collected by a Brilliance CT Big Bore system from Philips Healthcare (Philips Healthcare, Best, the Netherlands), with the following scanning conditions: bulb voltage at 120 kV, current ranging from 275 to 375 mA, slice thickness of 3.0 mm, and full resolution of 512 × 512. An injected contrast agent, iohexol, was used during the ceCT examination. Similarly, CT images from Sichuan Provincial People's Hospital, The First Affiliated Hospital of University of Science and Technology of China and Southern Medical University were acquired using a Somatom Definition AS 40 system from Siemens Healthcare (Siemens Healthcare, Forcheim, Germany), with the following conditions: bulb voltage ranging from 120 to140 kV, current ranging from 280 to 380 mA, slice thickness of 3.0 mm, and full resolution of 512 × 512. CT images from Daguan Hospital of Chengdu Jinjiang were acquired using a Somatom Definition AS 40 system from Siemens Healthcare (Siemens Healthcare, Forcheim, Germany), with the following conditions: bulb voltage 120 kV, current ranging from 200 to 250 mA, slice thickness of 2.5 mm, and full resolution of 512 × 512.
The dataset consists of clinically required non-contrast CT images (ncCT) and/or contrast CT images (ceCT) from patients with nasopharyngeal cancer before treatment.
- Training data will consist of CT images from 262 patients from five cohorts (150 paired CT, 32 ncCT and 108 ceCT) with a corresponding label map, as well as 500 unlabeled cases.
- Validation data will consist of 40 patients from external testing cohort: 20 with paired CT, 10 with only ncCT, and 10 with only ceCT.
- Testing data will consist of 100 patients from external testing cohort: 40 with paired CT, 30 with only ncCT, and 30 with only ceCT.
Note: All LN CTVs were annotated individually by oncologists using MIM Software and ITKSNAP, the annotation of each organ was also stored individually. The expected output from your algorithm should be a set of label maps.
Download
Please fill out the Registration form.
The training data (with labels) and validation data (without lables) can be downloaded at: GoogleDrive and BaiduNetDisk. The unzip password is segrap2023@uestc.
The training data (with labels) can be downloaded at: here,
and the unzip passowrd is lnctvseg@uestc.
The validation data (without labels) can be
downloaded at: GoogleDrive
and BaiduNetDisk.
A total of 500 unlabeled images are provided at: GoogleDrive and BaiduNetDisk. Participants may explore self-supervised or semi-supervised learning strategies to enhance model generalizability.
- Please fill out the EndUserAgreement, and email a scan of the signed document to segrap2025@163.com. After receiving online Registration form and EndUserAgreement, we will provide you with the unzip password for the Task02 validation set and Supplementary unlabeled data.
- SegRap2025 focuses on the GTV and LN CTV segmentation. Participants are encouraged to leverage OAR anatomical information to support GTV segmentation, but segmentation of OARs are not necessary.
- The use of foundation models is permitted, but additional external data are not allowed. Only the official data can be used.