Synopsis

Keywords: Data Analysis, Oxygenation

SvO₂ and OEF were estimated via susceptometry-based oximetry (SBO). The impact of user-defined ROI was investigated in this study. The results show that (1) the consistency of SvO₂ estimation depends on operators’ comprehension of MR-oximetry, (2) SvO₂ might be under-estimated with manual ROI selection, and (3) the operator-induced bias can be eliminated by the semi-automatic analysis. The preliminary data indicate the feasibility of the constructed semi-automatic analysis for consistent estimation of SvO₂ among operators and its potential for bulk analysis and clinical use.

INTRODUCTION

Cerebral oxygen extraction fraction (OEF), defined as the fractional amount of oxygen consumption over oxygen supply, can reflect the balance between cerebral blood flow (CBF) and oxygen consumption (CMRO₂). Accordingly, it has been considered as a potential biomarker for brain tissue viability and widely investigated in causality in a variety of diseases, such as Alzheimer’s disease¹ and metabolic disorders². In comparison to ¹⁵O PET and T₂-based oximetry, susceptometry-based oximetry (SBO), as a calibration-free approach without administration of radiation tracers, has become a great noninvasive alternative for measurement of venous oxygen saturation level (SvO₂) which can be subsequently used for OEF estimation³. The reliability and reproducibility of SvO₂ via SBO mainly depends on the accumulated phase of venous draining vessel relative to the surrounding brain tissue. However, errors might be additionally propagated into SvO₂ estimation in the analysis in terms of experienced and inexperienced operators. Specifically, ill-defined regions-of-interest (ROIs) of the draining vessel and the surrounding compartment might hamper estimation of phase modulation. In this study, we constructed a standardized semi-automatic analysis to eliminate the operator-induced bias in SvO₂ estimation. The inter-analyst agreement of SBO was investigated in SvO₂ in healthy subjects.

METHODS

Twenty healthy volunteers (13 males and 7 females, age mean 28.9±5.4 years) were enrolled. Based on SBO approach, SvO₂ was determined by measuring phase difference (∆𝜙) between intra-and extravascular compartments accrued over a time ∆𝑇𝐸, $$SvO_2=1-\frac{2|\Delta\phi|}{\gamma B_0\Delta\chi _{do} Hct(\cos^{2}\theta-1/3)\Delta TE}$$ where γ is the gyromagnetic ratio, Δχ_do represents the susceptibility difference between fully oxygenated and deoxygenated erythrocytes, Hct is the volume fraction of red blood cells, and θ is the vessel tilt angle relative to the main magnetic field (B₀).
A two-slice interleaved OxFlow sequence was conducted, which was originally developed to simultaneously quantify SvO₂ and whole-brain CBF^{4, 5}. Imaging was performed at 3.0T (Siemens Prisma) using a 64-channel head coil. The scanning parameters of OxFlow protocol were: 1x1x5 mm³ voxel size, flip angle of 15°, TR/TE1/ΔTE=20/5.5/3.71 ms in the superior sagittal sinus (SSS) for SvO₂ estimation. Image reconstruction and analysis were performed with in-house MATLAB (MathWorks, Natick, MA) scripts. To improve the consistency of SvO₂ estimation, in the semi-automatic post-processing, operators merely needed to locate the data for analysis and determine the ROI span via a rectangle comprising SSS. The SSS in the following was isolated by means of optimum global thresholds determined by maximizing the between-class variance⁶. The area surrounding the SSS (ref mask) was automatically segmented via morphological processing (i.e., dilation and erosion) of the SSS mask, generating an irregular shape of ten pixels length and four pixels above the SSS, with tiny vessel pixels excluded whose values are larger than the mean +3 standard deviations (Fig. 1A).
The operators comprised one experienced MR physicist (operator 1) with 5 years of practical experience in MR-based oximetry and one inexperienced trainee (operator 2) with more than one-year training period. The training for the latter included didactic lectures on MR physics and supervised OxFlow data analysis to get familiar with the in-house analysis pipeline. For comparison, free-hand ROIs for SSS and ref masks were also extracted by both operators. The measurement from the operator 1 was considered as the reference.
Descriptive data were presented as mean±SD. Wilcoxon signed-rank test and Spearman’s correlation were applied to analyze significance of differences in SvO₂ and OEF between manual and automatic ROI selections. Bland and Altman analysis was computed for comparison between operators. A probability value < 0.05 was considered as the criterion for statistical significance.

RESULTS

The ROIs (yellow contours) extracted by the semi-automatic analysis (Fig. 1A) and manually by the experienced operator (Fig. 1B) and the inexperienced operator (Fig. 1C) are overlaid on the SSS and the surrounding compartment in a representative magnitude image, showing that the ROIs are well-extracted by the constructed post-processing with the tiny vessel close to SSS excluded. Table 1 lists the SvO₂ and OEF estimated with semi-automatic and manual ROI selections. The mean SvO₂ is comparable in measurements by the experienced operator (left in Fig. 1D, P=0.39), regardless of the different ROI placements in the surrounding area (Fig. 1A and 1B), but significantly different between methods by operator 2 (right in Fig. 1D, P<0.001). Estimation with free-hand ROI shows smaller values of SvO₂ in operator 2 (Fig. 2A and left red box in 2C). On the other hand, parallel SvO₂ are found between operators when the semi-automatic analysis is applied (Fig. 2B and two right boxes in 2C). Extremely high correlation in SvO₂ between operators indicates the reliability and reproducibility of the semi-automatic analysis (data not shown, 𝜌 = 0.98, P <0.001). Figure 3 demonstrates that manual values are generally under-estimated in SvO₂ (𝜌=0.98, P<0.001, slope = 0.94).

CONCLUSIONS

The purpose of the study was to present a standardized semi-automatic analysis for elimination of operator-induced bias in SvO₂ estimation, which has direct impact on OEF assessment. Note that the measurements of SvO₂ by the experienced operator is well consistent with the previous study using a similar pulse sequence⁵ and considered as the reference. In conclusion, the results show a great inter-analyst agreement, indicating the standardized analysis can significantly maintain reliability and reproducibility of SvO₂ via SBO.

Acknowledgements

This work was supported by the National science and Technology Council (MOST 110-2314-B-110-005 -)