Related Experiment Video
Updated: Jun 15, 2025
Brain Infarct Segmentation and Registration on MRI or CT for Lesion-symptom Mapping
Published on: September 25, 2019
Comparison of neuroprognostic performance between manually and automatically computed gray-white matter ratios on
Chih-Hung Wang1, Chu-Lin Tsai2, Hua Li3
1Department of Emergency Medicine, College of Medicine, National Taiwan University, Taipei, Taiwan; Department of Emergency Medicine, National Taiwan University Hospital, Taipei, Taiwan; Department of Emergency Medicine, National Taiwan University Hospital Yunlin Branch, Douliu, Taiwan.
Aim:
To compare the prognostic accuracy of manually (mGWR) and automatically (aGWR) computed gray-white matter ratios on brain computed tomography for predicting neurological outcomes in adult post-cardiac arrest patients.
Methods:
We systematically searched the PubMed and Embase databases from their inception to August 2024. Studies providing sufficient data on mGWR or aGWR to predict neurological outcomes in adult post-cardiac arrest patients were selected. A Bayesian bivariate random-effects meta-analysis model was used to synthesize data. Between-study heterogeneity was quantified using the I² statistic, and publication bias was assessed with Deek's test.
Results:
A total of 42 studies, involving 8104 patients, were included in the meta-analysis (mGWR: 41 studies, 6843 patients; aGWR: 5 studies, 1261 patients; 4 studies reported both mGWR and aGWR). The pooled area under the curve (AUC) for mGWR was 0.77 (95 % credible interval [CrI], 0.73-0.81; I², 100 %), with a pooled sensitivity of 0.55 (95 % CrI, 0.50-0.61) and specificity of 0.96 (95 % CrI, 0.93-0.99). For aGWR, the pooled AUC was 0.84 (95 % CrI, 0.81-0.87; I², 98 %), with a pooled sensitivity of 0.53 (95 % CrI, 0.34-0.71) and specificity of 0.95 (95 % CrI, 0.88-0.99). Subgroup analyses did not identify the source causing the heterogeneity, including brain regions for GWR calculations, GWR calculation formulas, and threshold values. No significant publication bias was found (mGWR, p = 0.28; aGWR, p = 0.79).
Conclusions:
The neuroprognostic performance of mGWR and aGWR was comparable, with a slightly higher AUC for aGWR. aGWR shows potential as a standardized imaging biomarker for guiding treatment decisions.
