Dear BraTS team, I am using the BraTS-GLI dataset for a longitudinal project. I found that some subjects have multiple timepoints, for example BraTS-GLI-00005-100 and BraTS-GLI-00005-101, and some have 100/101/102/... I would like to ask whether the actual visit dates or scan intervals for these timepoints are available anywhere. I checked the metadata file, but I could not find scan dates, days from baseline, or follow-up intervals. Do these timepoint IDs only indicate scan order, or is there any file/resource that provides the actual timing information? Thank you very much. Best regards, Darui Xu

Created by David Xu xdr
**Title: Clarification on Shared Base IDs, Longitudinal Timepoints, and Patient-Level Data Splitting in BraTS 2024 GLI** Dear BraTS team, I am working with the BraTS 2024 GLI post-treatment dataset and would appreciate clarification regarding the interpretation of subject IDs that share the same base ID but have different suffixes. During my metadata and archive audit, I identified **608 base-ID groups containing multiple full subject IDs, comprising 1,599 subject records**. For example, the dataset includes cases such as: - `BraTS-GLI-00005-100` - `BraTS-GLI-00005-101` The supplementary metadata contains demographic and clinical fields, such as age, sex, glioma type, site, and site-specific subject ID. However, I could not find scan dates, time since treatment, or explicit information establishing the chronological relationship between these records. I understand that BraTS datasets may include longitudinal MRI examinations and that similar suffix conventions have been used in previous BraTS releases. However, I would like to confirm how these identifiers should be interpreted specifically in the **BraTS 2024 GLI post-treatment dataset**. Could the organizers please clarify the following? 1. **Patient identity:** Do records sharing the same base ID (e.g., `BraTS-GLI-00005`) definitively correspond to the same individual in the BraTS 2024 dataset? 2. **Temporal ordering:** Do suffixes such as `-100`, `-101`, `-102`, and `-103` represent sequential MRI examinations of the same patient? If so, does the numerical order reliably reflect chronological order? 3. **Longitudinal information:** Are actual scan dates, time intervals between examinations, or any mapping file linking these records available? In particular, is there a way to determine whether the scans were acquired before or after specific treatments, or how far apart they were acquired? 4. **Data leakage prevention:** For a machine-learning study, would you recommend keeping all records with the same base ID in the same training, validation, or test partition to prevent potential patient-level data leakage? Or are there cases in which a shared base ID does not imply the same patient? I am asking because the number of records in these groups is substantial, and excluding them entirely would unnecessarily reduce the available training data. If they represent repeated examinations of the same individuals, grouping them during data splitting would be preferable to discarding them. Any clarification or reference to the relevant dataset documentation would be greatly appreciated. Thank you for your time and for providing this valuable dataset.

Are actual visit dates or scan intervals available for BraTS-GLI longitudinal timepoints? page is loading…