Nevin Manimala Statistics

Evaluating the Impact of Intracranial Volume Correction Approaches on the Quantification of Intracranial Structures in MRI: A Systematic Analysis

J Magn Reson Imaging. 2023 Sep 13. doi: 10.1002/jmri.28974. Online ahead of print.


BACKGROUND: In neuroscience, accurately quantifying individual brain regions in large cohorts is a challenge. Differences in intracranial structures can suggest functional differences, but they also reflect the effects of other factors. However, there is currently no standardized method for the correction of intracranial structure measurements.

PURPOSE: To identify the optimal method to counteract the influence of total intracranial volume (TIV) and gender on the measurement of intracranial structures.

STUDY TYPE: Prospective.

POPULATION/SUBJECTS: One hundred forty-one healthy adult volunteers (70 male, mean age 21.8 ± 1.7 years).

FIELD STRENGTH/SEQUENCE: T1-weighted 3D gradient-echo sequence at 3.0 T.

ASSESSMENT: A radiologist with 5 years of work experience screened the raw images to exclude poor-quality images. Freesurfer then performed automated segmentation to obtain measurements of intracranial structures. Male-only, female-only, and TIV-matched sub-samples were created separately. Comparisons between the original data and these sub-samples were used to assess the effects of gender and TIV. Comparison the consistency between TIV-matched sample and corrected data that corrected by four methods: Proportion method, power-corrected proportion method, covariate regression method, and residual method.

STATISTICAL TESTS: Cohen’s d for examining group distribution disparities, t-tests for probing mean differences, correlation coefficients to assess the relationships between intracranial substructure measurements and TIV. Multiple comparison corrections were applied to the results.

RESULTS: The correlation coefficients between TIV and the volumes of intracranial structures ranged from 0.033 to 0.883, with an average of 0.467. Thirty significant volume differences were found among 36 structures in the original sample, while no differences were observed in the TIV-matched sample. Among the four correction methods, the residual method had highest consistency (similarity 94.4%) with the TIV-matched group.

DATA CONCLUSION: The variation in intracranial structure sizes between genders was largely attributable to TIV. The residual method offers a more accurate and effective approach for correcting the effects of TIV on intracranial structures.


PMID:37702125 | DOI:10.1002/jmri.28974

By Nevin Manimala

Portfolio Website for Nevin Manimala