To classify pathology in clinical brain MRIs of Neurofibromatosis Type 1 (NF1), for the purpose of creating NF1-specific brain growth charts. 

NF1 is an autosomal dominant disorder occurring in approximately 1 in 3000 births. The disorder is associated with nervous system tumors, macrocephaly, and increased rates of neurodevelopmental disorders including autism and ADHD. Thus, it is critical to understand NF1 brain growth trajectories and co-occurring pathology. However, prior work on NF1 brain growth is limited, particularly in early childhood. Clinically-acquired brain MRIs provide an untapped data resource to address these knowledge gaps.

Prior to generating brain growth charts from clinical MRIs, it is first necessary to annotate brain pathology that may affect quantitative image processing. We applied the following NF1-specific grading system to MRI radiology reports of NF1 patients at the Children’s Hospital of Philadelphia: Grade_0 (severe pathology that likely precludes automated image processing, e.g., extensive tumor); Grade_1 (intermediate pathology that likely does not interfere with image processing, e.g. small optic pathway glioma, <1cm largest dimension); Grade_2 (minimal/no pathology). Grade_1 sessions underwent manual annotation for specific pathologies. Anatomical MRIs from Grade_1 and Grade_2 sessions were processed with recon-all-clinical (FreeSurfer v7.4.1).

7,580 MRI sessions from 754 NF1 patients age 0-21y were graded (Grade_0, n=2646; Grade_1, n=3444; Grade_2, n=1490. Ongoing Grade_1 annotation completed on 90% of scans revealed common NF1 pathologies (optic pathway glioma, n=1765; possible but unconfirmed non-optic pathway intracranial gliomas, n=870; prominent ventricles, ventriculomegaly, n=295; sphenoid dysplasia, n=272). Median age-at-scan differed significantly across grades, likely reflecting longitudinal progression of pathology (Grade_0 median=10.9y; Grade_1 median=7.4y; Grade_2 median=4.7y; χ²=948.1, p<2.2x10^-16). MRIs from Grade_1 and Grade_2 sessions were successfully processed allowing for quantification of global and regional brain volumes.

This work establishes a foundation for developing NF1-specific growth charts from clinical brain MRIs, advancing our understanding of NF1 brain development and brain-behavior relationships.