Lesion‐network mapping for post‐stroke cognitive impairment

Abstract

AbstractBackgroundTo overcome difficulty obtaining functional neuroimaging data directly from acute stroke patients, researchers have developed a new mapping approach called the “indirect lesion‐network mapping.” This mapping approach uses a priori network information obtained from normative samples to account for brain areas functionally connected to patients’ local lesions. This new method has successfully mapped some peculiar stroke symptoms, such as aphasia and post‐stroke pain. However, it remains unclear whether this method can produce sensible mapping results for the post‐stroke cognitive impairment of community samples.MethodA total of 1,431 patients were enrolled from two university hospitals. They underwent a comprehensive neuropsychological test within one year post‐stroke. We used the Verbal Learning Test to examine memory performance and the trail‐making test‐A and digit symbol coding task to assess the speed of processing. We classified the patients into two groups— cognitively‐impaired (CI) (< ‐2.0 standard deviation) and cognitively‐unimpaired (CU) (> ‐1.0 SD)—based on the task performance. We calculated the lesion‐affected networks based on patients’ stroke lesions combined with the normative functional network data from the Human Connectome Project‐Aging cohort, n = 687. Then, we obtained and compared the lesion network maps for patients with versus without cognitive impairment.ResultA total of 348 patients showed impairment in the memory task, while 247 patients demonstrated impairment in the speed‐of‐processing task. For both tasks, the conventional lesion‐network mapping approach produced highly similar results between the CI versus CU groups—both groups were localized to the major hubs of the large‐scale functional networks, such as mid‐cingulate and insular cortices. However, the mapping results became more sensible when we contrasted the lesion‐network maps between the CI versus CU groups. For example, the memory impairment was localized to the functional network including the left medial and anterior temporal, inferior and dorsolateral frontal cortices, whereas the speed‐of‐processing impairment was localized to the network including both occipital and anterior temporal cortices.ConclusionWe suggest that some additional analysis steps to the conventional lesion‐network mapping, including the contrast analysis, can help improve the specificity and sensitivity of the results when targeting the stroke‐induced cognitive impairment of community samples.