Abstract
AbstractBackgroundChoroid Plexus (ChP) manual segmentation performance evaluation on non-contrast-enhanced MRI sequences compared to the gold-standard contrast-enhanced T1-w has never been done on a relevant cohort of subjects.PurposeTo investigate whether contrast-enhancing can be avoided when performing ChP manual segmentation. To select which non-contrast-enhanced sequence between T1-w and FLAIR could be used as a contrast-enhanced T1-w surrogate. To provide a quantification of the ChP volume error that non-contrast-enhanced sequences introduce.Materials and methodsSixty-one prospective Multiple Sclerosis patients were included in the study. ChP was separately segmented on T1-w, FLAIR, and contrast-enhanced T1-w sequences. Quantitative contrast metrics between the ChP and surrounding ventricles were calculated. Quantitative segmentation metrics were obtained using gold-standard segmentation as reference. To assess the spatial agreement between non-contrast-enhanced and contrast-enhanced sequences, the segmentations were non-linearly coregistered to the standard MNI152 space and the error distribution per slice was evaluated spanning axially and coronally.ResultsConcerning contrast metrics, ANOVA test revealed a statistically significant main effect between the sequences (pvalue<0.01). The post-hoc t-tests revealed higher Contrast-to-Noise-Ratio and Signal-to-Noise-Ratio for contrast-enhanced sequences than others. T1-w exhibits the lower Contrast-to-Noise-Ratio while Signal-to-Noise-Ratio was comparable between FLAIR and T1-w (mean Signal-to-Noise-Ratio/Contrast-to-Noise-Ratio: contrast-enhanced T1-w=23.77/18.49, T1-w=13.73/7.44, FLAIR=13.09/10.77). The segmentation metrics revealed that non-contrast-enhanced sequences had comparable Dice. FLAIR overestimated the ChP volume while T1-w introduced a lower bias (Percentage Volume Difference FLAIR:28.02±19.02%; T1-w:3.52±12.61%). The spatial variability analysis confirmed that ChP volume depiction presents spatial differences between segmentations. FLAIR generally underperformed T1-w.ConclusionThe quantitative analyses suggest that T1-w might be a good candidate as a surrogate of contrast-enhanced sequence for the ChP manual segmentation task to estimate ChP volume. On the contrary, FLAIR introduces a systematic overestimation bias.SummaryTo estimate the Choroid Plexus Volume with manual segmentation, contrast-enhanced T1-w can be replaced by non-contrast-enhanced T1-w because the quantified error is acceptable, while FLAIR overestimates the volume.Key Points-T1-w MRI sequence has the lowest contrast-to-noise ratio among the available sequences but provided lower error in evaluating the Choroid Plexus Volume compared to FLAIR, both looking at spatial and overall indices (Percentage Volume Difference=3.52±12.61%). FLAIR has a higher contrast-to-noise ratio than T1-w sequence but overestimates the Choroid Plexus Volume (Percentage Volume Difference=28.02±19.02%).-T1-w can be used as a surrogate of contrast-enhanced T1-w sequence in Choroid Plexus manual segmentation.Importance of the StudyIn this manuscript we present our experience concerning the use of non-contrast enhanced sequences when manually segmenting the Choroid plexus, from brain MRI. The quantification of Choroid plexus volume is becoming of great interest in recent years, however a direct comparison of gold-standard techniques based on contrast and sequences acquired without contrast agent is missing. Here we present this comparison with quantitative further analysis also on spatial pattern of manual segmentation obtained with the different sequences, suggesting a possible surrogate to the gold-standard sequence to be used in large cohort studies or to train artificial intelligence models.
Publisher
Cold Spring Harbor Laboratory