Simultaneous estimation of tongue volume and fat fraction using IDEAL-FSE

Abstract

Purpose: To determine whether high-resolution, high signal-to-noise ratio (SNR) images of the tongue acquired with IDEAL-FSE (iterative decomposition of water and fat with echo asymmetry and least squares estimation) will provide comparable volumetric measures to conventional nonfat-suppressed FSE imaging and to determine the feasibility of estimating the proportion of lingual fat in adults using IDEAL-FSE imaging.

Materials and methods: Healthy volunteers underwent magnetic resonance imaging of the tongue using both IDEAL-FSE and conventional FSE sequences. The tongue was manually outlined to derive both volumetric and fat fraction measures. Intraclass correlation coefficients (ICCs) were computed for intrarater measurement reliability and Spearman's rank correlation tested the relationship between IDEAL-FSE and conventional volumetric measures of the tongue.

Results: IDEAL-FSE imaging yielded almost identical volumetric measures to that of conventional FSE imaging in the same amount of scan time (IDEAL-FSE mean 64.1 cm(3); conventional mean 63.3 cm(3); r = 0.988, P < or = 0.01). The average fat signal fraction across participants was 26.5%. Intrarater reliability was excellent for all measures (ICC > or = 0.92).

Conclusion: Our results indicate that IDEAL-FSE provided similar lingual volume estimates to conventional FSE imaging obtained in both the current and previous studies. IDEAL-FSE measures of lingual fat composition may be useful in studies that aim to increase lingual muscle strength and volume in swallowing and speech-disordered populations.

(c) 2008 Wiley-Liss, Inc.

Figures

Figure 1

Schematic of intrinsic and extrinsic lingual muscles included in the outline for analysis. The dotted-outlined box indicates the aspects of muscles that were included in the outline.

Figure 2

T2-weighted IDEAL-FSE acquires separate fat fraction (a), fat (b) water (c), and recombined (d) images that were reconstructed using an online reconstruction algorithm, while conventional imaging only provides a nonfat-suppressed image (e). Manual outline of tongue shown in image (f).

Figure 3

Scatterplot shows lingual volumetric measures (Cartesian coordinates) from IDEAL-FSE and conventional images are highly correlated (r = 0.988; P ≤ 0.01).