Automatic Lateral Ventricle and Choroid Plexus Segmentation method in infant brain MR images

Motivation: Recently, volume changes of CP and LV have been studied in normal aging in relation to glymphatic system. However, such analyses are limited in pediatric imaging due to various obstacles such as age-related tissue intensity changes. Goal(s): Our goal is to develop a robust automatic segmentation method for LV and CP using noisy labels in infant brain MR images. Approach: We trained a supervised and weakly-supervised networks ensemble to segment LV to refine the noisy labels, and CP segmentation was conducted using the refined LV labels. Results: Our proposed method demonstrated improved performance in segmenting both LV and CP in infant brain images. Impact: Our CP and LV segmentation method provides improved performance in infant MRI , demonstrating the potential for more robust quantitative analysis in the infant population. This could help to explore the relationship between glymphatic functions and the early stage of neurodevelopment.