Norwegian Centre for Mental Disorders Research (NORMENT), Institute of Clinical Medicine, University of Oslo, Oslo, Norway; Division of Mental Health and Addiction, Oslo University Hospital, Oslo, Norway.
Norwegian Centre for Mental Disorders Research (NORMENT), Institute of Clinical Medicine, University of Oslo, Oslo, Norway; Department of Psychology, University of Oslo, Oslo, Norway; Division of Mental Health and Addiction, Oslo University Hospital, Oslo, Norway; Department of Psychiatry, University of Oxford, Warneford Hospital, Oxford, United Kingdom.
Biol Psychiatry Cogn Neurosci Neuroimaging. 2020 Dec;5(12):1095-1103. doi: 10.1016/j.bpsc.2020.06.014. Epub 2020 Jul 8.
Schizophrenia (SZ) and bipolar disorder (BD) share substantial neurodevelopmental components affecting brain maturation and architecture. This necessitates a dynamic lifespan perspective in which brain aberrations are inferred from deviations from expected lifespan trajectories. We applied machine learning to diffusion tensor imaging (DTI) indices of white matter structure and organization to estimate and compare brain age between patients with SZ, patients with BD, and healthy control (HC) subjects across 10 cohorts.
We trained 6 cross-validated models using different combinations of DTI data from 927 HC subjects (18-94 years of age) and applied the models to the test sets including 648 patients with SZ (18-66 years of age), 185 patients with BD (18-64 years of age), and 990 HC subjects (17-68 years of age), estimating the brain age for each participant. Group differences were assessed using linear models, accounting for age, sex, and scanner. A meta-analytic framework was applied to assess the heterogeneity and generalizability of the results.
Tenfold cross-validation revealed high accuracy for all models. Compared with HC subjects, the model including all feature sets significantly overestimated the age of patients with SZ (Cohen's d = -0.29) and patients with BD (Cohen's d = 0.18), with similar effects for the other models. The meta-analysis converged on the same findings. Fractional anisotropy-based models showed larger group differences than the models based on other DTI-derived metrics.
Brain age prediction based on DTI provides informative and robust proxies for brain white matter integrity. Our results further suggest that white matter aberrations in SZ and BD primarily consist of anatomically distributed deviations from expected lifespan trajectories that generalize across cohorts and scanners.
精神分裂症（SZ）和双相情感障碍（BD）都有大量影响大脑成熟和结构的神经发育成分。这需要一个动态的生命周期视角，即通过偏离预期寿命轨迹来推断大脑异常。我们应用机器学习对弥散张量成像（DTI）的白质结构和组织指标进行分析，以在 10 个队列中估计和比较 SZ 患者、BD 患者和健康对照（HC）受试者之间的大脑年龄。
我们使用来自 927 名 HC 受试者（年龄 18-94 岁）的不同 DTI 数据组合训练了 6 个交叉验证模型，并将模型应用于包括 648 名 SZ 患者（年龄 18-66 岁）、185 名 BD 患者（年龄 18-64 岁）和 990 名 HC 受试者（年龄 17-68 岁）的测试集，估计每个参与者的大脑年龄。使用线性模型评估组间差异，该模型考虑了年龄、性别和扫描仪。应用荟萃分析框架评估结果的异质性和可推广性。
10 倍交叉验证显示所有模型的准确率都很高。与 HC 受试者相比，包含所有特征集的模型显著高估了 SZ 患者（Cohen's d = -0.29）和 BD 患者（Cohen's d = 0.18）的年龄，其他模型也有类似的影响。荟萃分析得出了相同的结论。基于各向异性分数的模型比基于其他 DTI 衍生指标的模型显示出更大的组间差异。
基于 DTI 的大脑年龄预测为大脑白质完整性提供了信息丰富且稳健的替代指标。我们的研究结果进一步表明，SZ 和 BD 中的白质异常主要由偏离预期寿命轨迹的解剖分布差异组成，这种差异在队列和扫描仪之间具有普遍性。
