Corpus callosum in children with neurodevelopmental delay: MRI standard qualitative assessment versus automatic quantitative analysis

Abstract

Background: The corpus callosum (CC) is a key brain structure. In children with neurodevelopmental delay, we compared standard qualitative radiological assessments with an automatic quantitative tool.

Methods: We prospectively enrolled 73 children (46 males, 63.0%) with neurodevelopmental delay at single university hospital between September 2020 and September 2022. All of them underwent 1.5-T brain magnetic resonance imaging (MRI) including a magnetization-prepared 2 rapid acquisition gradient echoes - MP2RAGE sequence. Two radiologists blindly reviewed the images to classify qualitatively the CC into normal, hypoplasic, hyperplasic, and/or dysgenetic classes. An automatic tool (QuantiFIRE) was used to provide brain volumetry and T1 relaxometry automatically as well as deviations of those parameters compared with a healthy age-matched cohort. The MRI reference standard for CC volumetry was based on the Garel et al. study. Cohen κ statistics was used for interrater agreement. The radiologists and QuantiFIRE's diagnostic accuracy were compared with the reference standard using the Delong test.

Results: The CC was normal in 42 cases (57.5%), hypoplastic in 20 cases (27.4%), and hypertrophic in 11 cases (15.1%). T1 relaxometry values were abnormal in 26 children (35.6%); either abnormally high (18 cases, 24.6%) or low (8 cases, 11.0%). The interrater Cohen κ coefficient was 0.91. The diagnostic accuracy of the QuantiFIRE prototype was higher than that of the radiologists for hypoplastic and normal CC (p = 0.003 for both subgroups, Delong test).

Conclusions: An automated volumetric and relaxometric assessment can assist the evaluation of brain structure such as the CC, particularly in the case of subtle abnormalities.

Relevance statement: Automated brain MRI segmentation combined with statistical comparison to normal volume and T1 relaxometry values can be a useful diagnostic support tool for radiologists.

Key points: • Corpus callosum abnormality detection is challenging but clinically relevant. • Automated quantitative volumetric analysis had a higher diagnostic accuracy than that of visual appreciation of radiologists. • Quantitative T1 relaxometric analysis might help characterizing corpus callosum better.

Keywords: Brain; Child; Corpus callosum; Magnetic resonance imaging; Segmentation.