0851

Thalamic Lac/NAA threshold of 0.39 on proton MRS predicts outcome following therapeutic hypothermia in neonatal hypoxic ischemic encephalopathy

Subhabrata Mitra¹, Giles Kendal¹, Cristina Uria-Avellanal¹, Alan Bainbridge¹, Magdalena Sokolska¹, David Price¹, Xavier Golay², and Nikki Robertson¹

¹University College Hospital, London, United Kingdom, ²University College London, London, United Kingdom

Synopsis

Perinatal hypoxic ischemic brain injury remains a significant cause of neonatal morbidity and mortality. Accurate prediction of neurodevelopmental outcome following HIE is important for parental counselling and directing care towards optimised treatments. Proton MRS derived thalamic Lactate/N-acetylaspartate (Lac/NAA) peak area ratio at 1.5T was a robust MR outcome biomarker in the pre-cooling era; a cut off Lac/NAA of 0.29 accurately predicted outcome. The aim of this work is to identify the cut off threshold of thalamic Lac/NAA at 3T in the cooling era for outcome prediction.

Introduction

Perinatal hypoxic ischemic brain injury remains a significant cause of neonatal morbidity and mortality. Accurate prediction of neurodevelopmental outcome following HIE is important for parental counselling and directing care towards optimised treatments. Proton MRS derived thalamic Lactate/N-acetylaspartate (Lac/NAA) peak area ratio at 1.5T was a robust MR outcome biomarker(1) in the pre-cooling era; a cut off Lac/NAA of 0.29 accurately predicted outcome.

The aim of this work is to assess the predictive accuracy of thalamic Lac/NAA at 3T in the cooling era, using two spectrum fitting methods: AMARES(2) implemented in the jMRUi and TARQUIN(3) software packages and identify the cut off threshold of thalamic Lac/NAA at 3T in the cooling era for outcome prediction. Additional aim is tocompare thalamic Lac/NAA with other MR imaging biomarkers for outcome prediction.

Methods

Subjects and protocol: 55 term infants admitted to our institution with moderate to severe NE, who fulfilled criteria for therapeutic hypothermia, were included in the study. All infants underwent MRI and thalamic 1H MRS at 3T (PRESS, 64 averages, TR/TE = 2288ms/288ms, voxel size 15x15x15mm, scan time ∼7 min) between day 2-14 (median day 5).

MRS analysis: Acquired spectra (figure 1) were processed using both jMRUi and TARQUIN independently and Lac/NAA peak-area ratios were computed. With both fitting methods NAA and N-acetyl aspartylglutamate (NAAG) are fitted and combined for total NAA; for jMRUi a doublet is fitted for lactate and for TARQUIN both lactate and threonine are fitted at 1.3 ppm. For the final analysis, log10Lac/NAA was calculated as the data were skewed.

MRI analysis: Diffusion tensor imaging (DTI) data was acquired in all neonates (SE-EPI, TR/TE 8000/49, FOV:224x220, 2x2x2mm b=750 s/mm2, 32 directions). All datasets were corrected for eddy current distortions(4). Next, mean diffusivity (MD) and fractional anisotropy (FA) maps were generated(5). Next, regional MD and FA values were computed in regions of interest (ROI) generated automatically in the basal ganglia thalamic region (BGT) (figure 2). MR images were scored using modified Okereafor(9) and NICHD(10) scoring systems. Bayley III assessments were performed at 2 years; neurodevelopment outcomes were classified as normal with a composite score of 85 and above. Poor outcome was defined as death or a composite score <85.

Results

Lac/NAA cut off threshold of 0.39 on TARQUIN predicted motor outcome with 100% sensitivity and 97.4% specificity (AUC 0.997) (figure 3A), cognitive outcome with sensitivity 89.5% & specificity 97% (AUC 0.97) (figure 3C), language outcome with sensitivity 81% & specificity 96.7% (AUC 0.91) (figure 3D). Lac/NAA on jMRUi predicted motor outcome with similar sensitivity (100%) & specificity (94.3%) with a cut off threshold Lac/NAA of 0.3 (AUC 0.995) (figure 3B). BGT MD predicted motor outcome with a sensitivity 65% and specificity 100% (AUC 0.77). MRI scores correlated with log10Lac/NAA (figure 5A & 5B).

In summary, thalamic Lac/NAA at 3T is a robust predictor of outcome after HT following HIE. A cut off threshold of 0.39 for TARQUIN predicted outcome in all domains with high sensitivity and specificity at 2 yrs of age. Move over, thalamic Lac/NAA performed better than other MR biomarkers.

Acknowledgements

No acknowledgement found.

References

1.Thayyil et al., Pediatrics. 2010

2.Vanhamme et al., J Magn Reson. 1997

3.Wilson et al., Magn Reson Med. 2011

4.Andersson et al., Neuroimage. 2016

5. Basser et al., J Magn Reson B. 1994

6. http://cmictig.cs.ucl.ac.uk/wiki/index.php/Main_Page

7. http://brain-development.org/brain-atlases/

8. Cardoso et al., Neuroimage. 2013

9. Okereafor et al., Pediatrics. 2008

10. Shankaran et al., J Pediatr. 2015

Figures

Figure 1: Voxel placement for thalamic 1H MRS with representative spectra from neonates with normal (left) and abnormal (right) outcome following hypoxic ischaemic encephalopathy.

Figure 2. ROIs for MD and FA in BGT were defined automatically based on both an atlas based probabilistic tissue segmentation and on a joint multi-atlas label propagation and fusion of 50 neonatal brain regions (6-8).

Figure 3. Logistic regression analysis of log10Lac/NAA for outcome prediction. Lac/NAA threshold of 0.389 in TARQUIN gave 1 false positive (A) for motor outcome while a threshold of 0.3 in jMRUI gave 2 false positive results (B). Using threshold of 0.389 in TARQUIN, analysis for cognitive outcome reveals 1 false positive and 2 false negative (C), analysis for language outcome reveals 1 false positive and 4 false negatives (D).

Figure 4. log10Lac/NAA was significantly different between the normal and poor outcome groups in all three domains of outcome. Comparison of mean was performed using 1-way analysis of variance, presented with 95% confidence intervals (p<0.0001).

Figure 5. A & B presents the relationship between log10(Lac/NAA TARQUIN) with MRI scores. Log10(Lac/NAA TARQUIN) correlated with BGT MD (R2 0.47) (figure C).

Proc. Intl. Soc. Mag. Reson. Med. 26 (2018)

0851