Transcranial direct current stimulation (tDCS) is a noninvasive neuromodulation technique gaining more attention in neurodevelopmental disorders (NDDs). Due to the phenotypic heterogeneity of NDDs, tDCS is unlikely to be equally effective in all individuals. The present study aimed to establish neuroanatomical markers in typically developing (TD) individuals that may be used for the prediction of individual responses to tDCS. Fifty-seven male and female children received 2 mA anodal and sham tDCS, targeting the left dorsolateral prefrontal cortex (DLPFC_left), right inferior frontal gyrus, and bilateral temporoparietal junction. Response to tDCS was assessed based on task performance differences between anodal and sham tDCS in different neurocognitive tasks (N-back, flanker, Mooney faces detection, attentional emotional recognition task). Measures of cortical thickness (CT) and surface area (SA) were derived from 3 Tesla structural MRI scans. Associations between neuroanatomy and task performance were assessed using general linear models (GLM). Machine learning (ML) algorithms were employed to predict responses to tDCS. Vertex-wise estimates of SA were more closely linked to differences in task performance than measures of CT. Across ML algorithms, highest accuracies were observed for the prediction of N-back task performance differences following stimulation of the DLPFC_left, where 65% of behavioral variance was explained by variability in SA. Lower accuracies were observed for all other tasks and stimulated regions. This suggests that it may be possible to predict individual responses to tDCS for some behavioral measures and target regions. In the future, these models might be extended to predict treatment outcome in individuals with NDDs.

经颅直流电刺激（tDCS）是一种无创神经调节技术，在神经发育障碍（NDD）中受到越来越多的关注。由于 NDD 的表型异质性，tDCS 不可能对所有个体都同样有效。本研究旨在建立典型发育 (TD) 个体的神经解剖学标记，可用于预测个体对 tDCS 的反应。 57 名男性和女性儿童接受了 2 mA 阳极和假 tDCS，针对左侧背外侧前额叶皮层（DLPFC_左侧）、右侧额下回和双侧颞顶交界处。根据不同神经认知任务（ N -back、侧翼、穆尼面部检测、注意力情绪识别任务）中阳极和假 tDCS 之间的任务表现差异来评估对 tDCS 的反应。皮质厚度 (CT) 和表面积 (SA) 的测量来自 3 特斯拉结构 MRI 扫描。使用一般线性模型（GLM）评估神经解剖学和任务表现之间的关联。采用机器学习 (ML) 算法来预测对 tDCS 的反应。 SA 的顶点估计与任务绩效差异的联系比 CT 的测量更密切。在 ML 算法中，在刺激 DLPFC_左侧后N背任务表现差异的预测中观察到了最高的准确度，其中 65% 的行为差异是由 SA 的变异性解释的。所有其他任务和刺激区域的准确度均较低。这表明有可能预测个体对某些行为测量和目标区域对 tDCS 的反应。 将来，这些模型可能会扩展到预测 NDD 患者的治疗结果。