Fig 1.
Composition of the analyzed cohort of patients.
(AF: atrial fibrillation, MRI: magnetic resonance imaging, N: number).
Table 1.
Clinical characteristics of included patients.
(AF: atrial fibrillation; NIHSSS: National Institute of Health Stroke Scale Score; IQR: Interquartile range; n: number; SD: standard deviation).
Fig 2.
Voxelwise stroke lesion counts for the whole cohort and the diagnostic subgroups of different AF status.
(A): Voxelwise stroke lesion counts and the margins of 1% and 4% total empirical stroke probability for the entire cohort of 582 acute ischemic stroke patients. Note the accumulation of ischemic lesions in the right insula, as illustrated by the 4% stroke probability margin. (Red-to-yellow: stroke lesion counts scaled between 1 to 20 lesions per voxel; green: margin of 1% total empirical stroke probability; blue: margin of 4% total empirical stroke probability). (B): Voxelwise empirical stroke probabilities for patients with new (top), known (middle) and without (bottom) AF. Smaller infarct volumes in patients without AF (cf. Fig 3) lead to less overlap of the ischemic lesions across subjects and thus much fewer voxels above the 4% empirical stroke probability threshold (set to display for illustration) while for the larger strokes of new and known AF ischemic lesions accumulate in both middle cerebral artery territories and centrally around the insula, in particular. (C): Voxelwise empirical stroke probabilities (thresholded at 10%) reveal that AF (new and known combined) leads to an accumulation of stroke lesions in the right MCA territory and insula, in particular. (AF: atrial fibrillation; MCA: middle cerebral artery).
Fig 3.
Cubic root total infarct volume by AF status (no, new and known AF).
Fig 4.
Association of stroke size and voxelwise ischemic involvement across 582 patients.
Larger strokes were associated with more frequent ischemic lesions in the MCA territory of both hemispheres and parts of the posterior circulation while smaller strokes were not associated with more frequent ischemic insults of any brain area. (Red-to-yellow: areas significantly associated with higher crIV, blue-to-lightblue: areas significantly associated with lower crIV across all n = 582 patients; thresholded at FDR-corrected q = 0.01; age and gender included as covariates of no interest, AF status unmodeled).
Fig 5.
Differences of infarct locations: New AF versus no AF.
(A): New AF was, compared to patients without AF, associated with ischemic lesions in a) the right (peri-) insular cortex and the caput of the right caudate nucleus and b) in the left PCA territory. Patients without AF had significantly more often strokes in the left parietal lobe than patients with new AF. (Red-to-yellow: areas significantly more often associated with strokes in patients with new AF vs. without AF, blue-to-lightblue: vice versa and green: overlap projections; thresholded at FDR-corrected q = 0.01; age and gender were included as covariates of no interest, with infarct volume unaccounted for in this analysis). (B): Including infarct volume into the analysis revealed that no specific brain region was associated with significantly more stroke lesions in patients with new AF compared to patients without AF. Vice versa, patients without AF still had more often left parietal involvement compared to patients with new AF. (Red-to-yellow: areas significantly more often associated with strokes in patients with new AF vs. without AF [none], blue-to-lightblue: vice versa; FDR-corrected q = 0.01; with removal of the confounding effect of cubic root infarct volume normalized for brain size, age and gender).
Table 2.
Tabulated stroke areas that significantly differed in the comparisons between diagnostic groups (without correcting for infarct volume, BSGLMM-VLSM, FDR-corrected q = 0.01 within 1% of total empirical stroke probabilities).
Fig 6.
Differences of infarct locations: New AF versus known AF.
(A): New AF was associated with more strokes in the central territory of the right MCA than known AF. The reverse comparison of known AF versus new AF did not yield a significant difference. (Red-to-yellow: areas significantly more often associated with strokes in new AF vs. known AF patients; blue-to-lightblue: vice versa; FDR-corrected q = 0.01; age and gender were included as confounds, with infarct volume unaccounted for in this analysis). (B): New AF was also more often associated with right central MCA-strokes compared to known AF after including infarct volume as a confound. Here, the left parietal lobe was significantly less often affected in new AF compared to known AF. (Red-to-yellow: areas where new AF was associated with significantly more strokes than in patients with known AF; blue-to-lightblue: vice versa; FDR-corrected q = 0.01; with removal of the confounding effect of cubic root infarct volume, age and gender).