Abstract
We developed a cortical language map from performance data on a language test battery in patients with brain lesions. The research problem was how to select the subtest that was most related to the function of each cortical area from the battery. When studied by voxel-based lesion-symptom mapping (VLSM), patients were divided into two groups: those with and without a lesion at each particular region. We considered the task that optimally discriminated between the two groups to be the task most related to the function of a given region. One hundred and fifty left-lesioned patients were examined using the Japanese Standard Language Test of Aphasia (SLTA), which is composed of 26 subtests. Using logistic discriminant analysis, we selected the subtest that optimally discriminated the lesioned and non-lesioned groups for each cortical region. Patients with left middle frontal gyrus (area 46) lesions were optimally discriminated from patients without lesions in that area by the speech sound-kana letter choice matching subtest. Patients with lesions in the inferior postcentral gyrus were optimally distinguished by the disturbance of word repetition. Patients with lesions in the anterior cingulate gyrus were characterized by impaired performance on the category fluency subtest. Voxel-based discriminant analysis can thus select the subtest that can be regarded as most related to the function of each cortical area.