According to the paper's abstract, multi-parametric MRI is the gold standard for neuro-oncology, yet acquiring a complete four-modality protocol can be unfeasible in real-world clinical practice. This creates a challenge for automated systems that rely on all modalities for brain tumor segmentation.
To address this, the authors propose CAPS (Clinically-Verifiable Anatomy-Prioritized Synthesis), a method that synthesizes missing MRI modalities with an emphasis on clinically verifiable anatomy. The abstract notes that existing automated systems 'force a' particular handling of missing modalities, though the full details are not included in the available text.
CAPS is presented as a way to make segmentation more robust when some MRI sequences are absent. The paper is available on arXiv under identifier 2610.28179.