Framework for Categorizing Intracerebral Hemorrhage Age: A Step Towards Fully-Automated Characterization and Visualization

Previous deep learning networks have autonomously identified, segmented, and quantified hematoma volumes in MR images of intracerebral hemorrhages (ICH). This information derived from this analysis would periodically augment surgical decisions during minimally invasive ICH evacuations. A limitation of these autonomous processes is the MRI contrast variations with varying clot ages precludes a generalizable CNN for ICH. We propose a multiparametric image processing pipeline for categorizing clots on the basis of their age, as determined by presented image contrast relative to local white matter. This determination can be used to select a properly trained CNN model based on age classification.