Hello, I would like to clarify something regarding the structure of the BraTS 2023 Adult Glioma (GLI) training dataset, particularly the presence of multiple scans from the same patient. I noticed that some cases have identical patient identifiers but different suffixes. For example: - BraTS-GLI-00008-000 - BraTS-GLI-00008-001 Based on a previous discussion here, I understand that these suffixes represent acquisitions at different timepoints from the same patient. I am asking because I recently received a reviewer comment concerning potential "patient leakage" in my brain tumor segmentation study using BraTS 2023. Interestingly, this aspect of the dataset does not seem to be explicitly addressed in many of the research papers I have consulted. I would therefore appreciate clarification from the dataset organizers or researchers familiar with its construction. I have a few questions: 1. When developing segmentation models, should all scans belonging to the same patient be assigned to the same training, validation, or test partition? In other words, is patient-level splitting the recommended practice to avoid potential data leakage? 2. Were the official BraTS 2023 training/validation/test partitions constructed with patient-level separation, ensuring that different timepoints from the same patient cannot appear in different partitions? 4. Is there an official list or metadata file identifying which cases belong to the same patient, or should researchers rely on the case naming convention to establish these groups? Thank you!