PLoS ONE (Jan 2015)

Lesion Location-Based Prediction of Visual Field Improvement after Cerebral Infarction.

  • Bum Joon Kim,
  • Yong-Hwan Kim,
  • Namkug Kim,
  • Sun U Kwon,
  • Sang Joon Kim,
  • Jong S Kim,
  • Dong-Wha Kang

DOI
https://doi.org/10.1371/journal.pone.0143882
Journal volume & issue
Vol. 10, no. 11
p. e0143882

Abstract

Read online

Although the prognosis of ischemic stroke is highly dependent on the lesion location, it has rarely been quantitatively utilized. We investigated the usefulness of regional extent of ischemic lesion (rEIL) predicting the improvement of visual field defect (VFD) in patients with posterior cerebral artery infarction.The rEILs were measured in each individual cortex after transforming the lesions to a standard atlas. Significant improvement of VFD was tentatively defined as 20% improvement at 3 months after stroke. The performances of clinical and imaging variables predicting significant improvement were measured by support vector machine. The maximum performance of variables predicting the significant improvement was compared between subgroups of variables (clinical, baseline severity and lesion volume) and the effect of adding rEIL to those subgroups of variables was evaluated.A total of 35 patients were enrolled in this study. Left PCA infarct, MR-time from onset, rEILs in the lingual, calcarine, and cuneus cortices were good prognostic indicators of hemi-VFD (performance for predicting the significant improvement: 72.8±11.8%, 66.1±11.2%, respectively). A combination of the rEILs of each cortical subregions demonstrated a better predictive performance for hemi-VFD (83.8±9.5%) compared to a combination of clinical variables (72.8±11.8; p<0.001), baseline severity (63.0±11.9%; p<0.001), or lesion volume (62.6±12.7%; p<0.001). Adding a rEIL to other variables improved the prognostic prediction for hemi-VFD (74.4±11.6% to 91.3±7.7%; p<0.001).An estimation of rEIL provides useful information regarding the ischemic lesion location. rEIL accurately predicts the significant improvement of VFD and enhances the prediction power when combined with other variables.