Segmentation of brain metastases using deep learning with missing MRI sequences: a multicenter study

Takeaway

  • In people with brain metastases, the input-level dropout (ILD) model detected metastatic voxels in the brain with equivalent accuracy to that of the DeepLab V3 model while performing significantly better at segmentation and generating a lower false positive rate.

Why this matters

    Recent studies have demonstrated the potential of using artificial intelligence for analysis of magnetic resonance imaging (MRI) data in people with brain metastases. However, a deep learning (DL) model for detection and segmentation of brain metastases that provides generalizability and clinical utility in the midst of missing MRI data is yet to be identified.