Topographical Association Between Left Ventricular Strain and Brain Lesions in Patients With Acute Ischemic Stroke and Normal Cardiac Function

Figure 1. Flow diagram of patient selection. *One of the following: (1) significant stenosis confirmed by angiography; or (2) angiography was not performed, but echocardiography showed regional wall motion abnormalities consistent with ischemic heart disease. †One of the following: (1) diagnosed within 6 months; (2) any treatment for cancer in the previous 6 months; (3) recurrence; or (4) distant metastasis. ‡Lesion drawing and volume measurement were performed using an in‐house machine learning‐based program. §A total of 68 patients were excluded due to voxels with lesion overlap below the minimum cutoff value. LVEF indicates left ventricular ejection fraction; MRI, magnetic resonance imaging; SMC, Samsung Medical Center; and SVR‐LSM, support vector regression lesion‐symptom mapping.

Figure 2. SVR‐LSM analysis to identify the location of brain lesions topographically associated with impaired LV GLS. A, Simple lesion overlap topography of all 286 patients. B, Unthresholded SVR‐β map includes only voxels with 10 or more overlapping lesions. This map shows how strongly each voxel contributes to the overall SVR‐LSM model, allowing for the identification of areas that may be important in symptom prediction. C, The voxel‐wise thresholded z(p) values measure the statistical significance of the SVR‐β values for each voxel, which is determined by thresholding the voxels at a P value of 0.05 and then performing cluster‐level correlations with 1000 permutations (cluster threshold of P<0.05). Age and sex were regressed on both lesion data and behavioral score data as covariates. GLS indicates global longitudinal strain; Lt, left; LV, left ventricle; Rt, right; SVR, support vector regression; and SVR‐LSM, support vector regression lesion‐symptom mapping.

Figure 3. A heatmap with a dendrogram indicating the clustering of impaired RLS in 18 LV segments along the coronary arterial territories and LV longitudinal axis. The K‐means cluster analysis was performed to divide the 18 impaired RLS into 3 groups according to the coronary arterial territories (LAD, LCX, and RCA) and LV longitudinal topography (apical, mid, and basal segments). The dendrogram represents the hierarchical clustering of groups obtained by K‐means cluster analysis. Both rows and columns of the heatmap represent 18 impaired RLS, and the heatmap shows the correlation among them. The strains clustered well in the apical segment of the LV, whereas a clustering pattern along the coronary arterial territories was not evident. This analysis suggested that the distribution pattern of impaired RLS was associated with LV longitudinal topography, which was independent of coronary arterial territory. LAD indicates left anterior descending artery; LCX, left circumflex artery; LV, left ventricle; RCA, right coronary artery; and RLS, regional longitudinal strain.

Figure 4. Topographically associated brain lesions with impaired RLS for each LV segment. RLS of the anterior, anteroseptal, anterolateral, inferior, inferoseptal, and inferolateral walls was calculated by averaging the strains at the apical (red), midventricular (orange), and basal (yellow) measures in each myocardial wall. Regional longitudinal strain showed a topographical association with right insula and left parietal cortical lesions. In particular, as the LV segment with impaired RLS shifted from the apex to the basal segment, the location of the associated brain lesions tends to move from the rostral to caudal region of the right insula and from the parietal to temporal region for left hemisphere. LV indicates left ventricle; and RLS, regional longitudinal strain.