Purpose: Children treated for brain tumors are at an increased risk for cognitive impairments due to the effect of radiation therapy on developing white matter (WM). Although damage to long-range WM is well documented in pediatric brain tumor survivors, the effect of radiation therapy on short-range WM remains unelucidated. We sought to clarify whether radiation treatment affects short-range WM by completing a virtual dissection of these connections and comparing their microstructural properties between brain tumor survivors and typically developing children.
Methods and materials: T1-weighted and diffusion-weighted magnetic resonance images were acquired for 26 brain tumor survivors and 26 typically developing children. Short-range WM was delineated using a novel, whole-brain approach. A random forest classifier was used to identify short-range connections with the largest differences in microstructure between patients and typically developing children.
Results: The random forest classifier identified differences in short-range WM microstructure across the brain with an accuracy of 87.5%. Nine connections showed the largest differences in short-range WM between patients and typically developing children. For these connections, fractional anisotropy and axial diffusivity were significantly lower in patients. Short-range connections in the posterior fossa were disproportionately affected, suggesting that greater radiation exposure to the posterior fossa was more injurious to short-range WM. Lower craniospinal radiation dose did not predict reduced toxicity to short-range WM.
Conclusions: Our findings indicate that treatment for medulloblastoma resulted in changes in short-range WM in patients. Lower fractional anisotropy and axial diffusivity may reflect altered microstructural organization and coherence of these connections, especially in the posterior fossa. Short-range WM may be especially sensitive to the effect of craniospinal radiation therapy, resulting in compromise to these connections regardless of dose.
Copyright © 2023 Elsevier Inc. All rights reserved.