BACKGROUND Late-life depression (LLD) and mild cognitive impairment (MCI) are risk factors for Alzheimer disease (AD). However, the interactive effect between LLD and MCI in the progression to AD remains unknown. The purpose of this research is to clarify whether this interaction exists and determined the characteristics of the structural change patterns in LLD and MCI. METHOD To address this question, a total 225 participants (91 with intact cognitive function (IC), 34 with MCI, 35 with LLD-IC, 47 with LLD-MCI and 18 with AD) were recruited for the current study and their T1 scanning were acquired. Machine learning was applied to estimate the brain's age gap according to grey matter information (thickness and volume was calculated based on the Human Connectome Project Multi-Modal Parcellation version 1.0 and the Desikan atlas). A structural covariance network (SCN) was constructed based on grey matter volume. Rich-club analysis, global network properties and the Jaccard distance were utilized to describe the topological features in each cohort. Their cognitive functions (executive function, processing speed and memory) were evaluated by a full-scale battery of neuropsychological tests. RESULT The interactive effect between LLD and MCI was detected through the brain age gap. The estimated age was positively correlated with processing speed and memory in LLD and non-LLD subjects. In the SCN analysis, the rich-club coefficient and global network properties were disrupted in the MCI group, but remained normal in the LLD-IC, LLD-MCI and AD groups. There was a significant discrepancy in brain structural change patterns between the AD and other cohorts by the Jaccard distance. CONCLUSION The application of machine learning reflects that synergies between LLD and MCI could increase the risk of developing AD. According to the SCN, the structural coordination was disrupted in MCI and was kept normal in the other cohorts, while the discrepancies in brain structural change patterns appeared in AD. Overall, the brain age gap could be a potential predictor of AD, and the Jaccard distance has the potential to be a new type of SCN analysis indicator.

中文翻译：

---

根据机器学习和结构协方差网络分析确定LLD和MCI对大脑衰退的影响。

背景技术晚期抑郁症（LLD）和轻度认知障碍（MCI）是阿尔茨海默病（AD）的危险因素。但是，LLD和MCI之间的相互作用在AD进展中的作用仍然未知。这项研究的目的是弄清这种相互作用是否存在，并确定了LLD和MCI中结构变化模式的特征。方法为了解决这个问题，共招募了225名参与者（91名具有完整的认知功能（IC），34名MCI，35名LLD-IC，47名LLD-MCI和18名AD）参加本研究并进行T1扫描被收购。应用机器学习根据灰质信息估计大脑的年龄差距（厚度和体积是根据“人类Connectome项目多模态融合1.0版和Desikan图集计算的”）。基于灰质体积构建了结构协方差网络（SCN）。利用Rich-club分析，全局网络属性和Jaccard距离来描述每个队列的拓扑特征。他们的认知功能（执行功能，处理速度和记忆力）通过一系列全面的神经心理学测试进行评估。结果通过脑年龄差异检测了LLD和MCI之间的相互作用。在LLD和非LLD受试者中，估计年龄与处理速度和记忆力呈正相关。在SCN分析中，MCI组的丰富俱乐部系数和全局网络属性被破坏，而LLD-IC，LLD-MCI和AD组则保持正常。根据雅克德距离，AD和其他队列之间的大脑结构变化模式存在显着差异。结论机器学习的应用反映了LLD和MCI之间的协同作用可能会增加罹患AD的风险。根据SCN的研究，MCI中的结构协调被破坏，而其他队列中的结构协调则保持正常，而AD中则出现了脑结构变化模式的差异。总体而言，脑年龄差距可能是AD的潜在预测指标，而Jaccard距离也有可能成为新型SCN分析指标。