Brain metastases are the most common form of CNS malignancy in adults, however their evaluation in clinical setting often proves challenging. This difficulty largely arises from the frequent occurrence of multiple small metastases in one patient. Moreover, detailed analysis of multiple lesions across serial scans is challenging due to the substantial time required to evaluate each metastasis across studies. Therefore, development of automated segmentation tools for brain metastases is crucial for maintaining a high level of patient care. Accurate detection of small metastatic lesions, particularly those smaller than 5 mm is essential for patient prognosis, as missing even one lesion can lead to repeated interventions and treatment delays. Furthermore, the gross total volume of brain metastases in a patient is an important predictor of patient outcomes, which is currently unavailable in clinical practice due to the absence of translatable volumetric segmentation tools.

​

       Addressing this issue requires the development of novel segmentation algorithms that can detect and accurately volumetrically segment all lesions. Existing algorithms, including the nnUnet, show high dice scores for larger metastases but may miss or perform poorly on small metastases. The BraTS Brain Metastases Challenge is essential for development of advanced segmentation and detection algorithms for brain metastases, which can be easily incorporated into clinical practice. The objective is to identify segmentation algorithms capable of segmenting both large and small metastases on diagnostic MRI, using T1, T1 post contrast, T2, and FLAIR sequences. This will provide standardized autosegmentation algorithms that will be open source and available to institutions for incorporation into their clinical and research workflows.