Subcortical brain segmentation in 5-year-olds: validation of FSL-FIRST and FreeSurfer against manual segmentation
Lidauer, Kristian (2021-01-12)
Subcortical brain segmentation in 5-year-olds: validation of FSL-FIRST and FreeSurfer against manual segmentation
Lidauer, Kristian
(12.01.2021)
Julkaisu on tekijänoikeussäännösten alainen. Teosta voi lukea ja tulostaa henkilökohtaista käyttöä varten. Käyttö kaupallisiin tarkoituksiin on kielletty.
suljettu
Julkaisun pysyvä osoite on:
https://urn.fi/URN:NBN:fi-fe202102266087
https://urn.fi/URN:NBN:fi-fe202102266087
Tiivistelmä
Recent developments in magnetic resonance imaging have facilitated the paediatric neurodevelopmental studies. Developing accurate subcortical volumetric quantification tools is a crucial issue for the field, as they could potentially eliminate the need for challenging and highly time-consuming manual segmentation, which is currently considered the gold standard in volumetric segmentation. In this study the accuracy of two automated segmentation tools, FSL-FIRST with three different boundary correction settings and FreeSurfer is compared against manual segmentation on subcortical nuclei, including the hippocampus, amygdala, thalamus, putamen, globus pallidus, caudate and nucleus accumbens. To determine the accuracy and consistency of these automated methods, volumetric and correlation analyses were computed against manual segmentation in 5-year-old children (N=80).
The results show that both FIRST and FreeSurfer overestimate the volume on all structures except the caudate. The results also indicate that the accuracy of the automated methods is dependent of the structure. Small structures such as the amygdala and nucleus accumbens, which are visually difficult to distinguish, produced considerably overestimations and weaker correlations with all automated methods, whereas larger and more readily distinguishable structures such as the caudate and putamen produced notably lower overestimations and stronger correlations. Overall, the segmentations performed by FIRST’s Default pipeline were the most accurate, while FreeSurfer’s results were weaker across the structures.
In line with prior studies ,these results suggest that the accuracy of automated segmentation tools is not perfect with respect to manually defined structures. However, apart from amygdala and nucleus accumbens, FSL FIRST’s agreement could be considered satisfactory (r > 0,74 and ICC > 0,68 with highest values for the striatal structures (putamen, globus pallidus and caudate nucleus) (r > 0,77 and ICC > 0,87).
The results show that both FIRST and FreeSurfer overestimate the volume on all structures except the caudate. The results also indicate that the accuracy of the automated methods is dependent of the structure. Small structures such as the amygdala and nucleus accumbens, which are visually difficult to distinguish, produced considerably overestimations and weaker correlations with all automated methods, whereas larger and more readily distinguishable structures such as the caudate and putamen produced notably lower overestimations and stronger correlations. Overall, the segmentations performed by FIRST’s Default pipeline were the most accurate, while FreeSurfer’s results were weaker across the structures.
In line with prior studies ,these results suggest that the accuracy of automated segmentation tools is not perfect with respect to manually defined structures. However, apart from amygdala and nucleus accumbens, FSL FIRST’s agreement could be considered satisfactory (r > 0,74 and ICC > 0,68 with highest values for the striatal structures (putamen, globus pallidus and caudate nucleus) (r > 0,77 and ICC > 0,87).