Can we explain machine learning-based prediction for rupture status assessments of intracranial aneurysms?