While acknowledging the hardships caused by COVID-19, the pandemic also provided a unique opportunity to study mental well-being and individual vulnerability or resilience.^{1, 2} Sociodemographic, psychological factors, and lifestyles, have been identified as predictors of mental health during COVID-19.³ Our previous study demonstrated the relevance of the interplay between psychological measures and brain networks' functional connectivity (FC).⁴ However, important questions remain to be addressed. For example, can FC—alone or in combination with other measures – predict longer-term mental health? Additionally, most studies focus on emotional aspects (psychological distress), although mental health (MH) comprises emotional, psychological (personal growth, [PG]), and social (loneliness) well-being components, which were differently impacted during the pandemic.³ This study aims to investigate if there exists specificity between FC measures and long-term changes across MH components, knowing the links between brain networks and ‘resilience processes’.^{5, 6}

We studied 702 healthy, middle-aged individuals (350 women, age: 50.66 ± 6.98 years) who met the criteria in a 2023 study by Bayes-Marin and colleagues.³ All participants gave written informed consent according to the Declaration of Helsinki. The study protocol was approved by the Comitè Ètic d'Investigació de la Fundació Unió Catalana d'Hospitals (CEIC-17/06). Resting-state functional magnetic resonance imaging images acquired before the COVID-19 outbreak were preprocessed, and system segregation (SyS; integration-segregation balance) was calculated for seven resting state networks (RSN) [see⁴] and Data S1). Multinomial logistic regressions were fitted to predict trajectory membership for the three MH components (Resilient or Chronic trajectories, for psychological distress and loneliness, and Resilient, Progressively Ascending, or Worsening for PG, as captured by growth mixture models contrasting pre- versus during-pandemic observations within two-year follow-up (see 3, Data S1 and Fig. 1a). RSN models included FC for seven RSNs. Full models combined significant RSN measures and significant predictors found in our previous study (age, sex, monthly income, stress coping, personality, general health, and lifestyle habits) (see³). Non-RSN models were as full models but without RSN data, and through likelihood ratio tests (non-RSN vs full models) we assessed whether the goodness of fit improved by adding FC measures to sociodemographic, psychological, and lifestyle measures. In total, we fitted three models (RSN, non-RSN, and full) for each of the three MH components.

The emotional trajectory membership (Fig. 1a) was significantly predicted by the Salience Network (SN) FC (Fig. 1b), revealing that a more functionally integrated SN (i.e., lower SyS) was more representative of Resilient trajectories in comparison to Chronic and marginally to Moderate-Resilient ones. In the full model, the significance of SN-FC as a predictor was reduced, and the comparison of full versus non-RSN models was non-significant (Fig. 1c).

The psychological trajectory membership (Fig. 1a) was significantly predicted by the Dorsal Attention Network (DAN), and marginally by Limbic Network (LN) and Fronto-Parietal Network (FPN). However, in the full model, only FPN-FC remained strictly significant, with a higher probability of belonging to the resilient trajectory with a greater integrated FPN. Notably, the full model was significantly better than the non-RSN model (Fig. 1c).

Finally, social trajectory membership was not significantly predicted by any of the RSN-SyS values (R² = 0.008).

Our findings indicate that measures of FC reflecting the integration-segregation of principal RSNs offer distinct predictions for long-term MH outcomes across the COVID-19 pandemic. This study suggests that pandemic anxious-depressive trends were affected by SN-SyS. However, emotional trajectories were predicted by a simpler and equally informative model that did not include RSN information, suggesting that lifestyles and psychological factors are enough to describe them. Nevertheless, when studying PG, baseline FPN-SyS was shown to add meaningful information to the model derived from aggregated sociodemographic, psychological factors, and lifestyles. Individual differences in PG maintenance likely reflect the capacity to thrive through reappraising and attaching value to stressful situations.⁷ As such, the associations found between PG and FPN connectivity, commonly linked to cognitive flexibility and control processes,⁸ may reveal the importance of cognition within psychological aspects. Finally, not finding any RSN-SyS associated with social well-being may be due to measuring the individual's subjective perception but not the direct engagement in social contacts, and/or the previously reported paradoxical effects of the outbreak on loneliness during the outbreak.⁹ Altogether, the present findings contribute to our previous observations,⁴ revealing the impact of the DMN- and FPN-SyS on emotional trajectories through stress-perception. This supports the triple-network perspective.^{5, 6, 10}

Overall, our findings suggest that assessing brain network integration versus segregation aids in predicting individual resilience and vulnerability across MH dimensions, allowing for early identification of at-risk individuals, and the design and evaluation of personalized preventive strategies.

在承认 COVID-19 造成的困难的同时，这场大流行也提供了研究心理健康和个人脆弱性或复原力的独特机会。^{1, 2}社会人口学、心理因素和生活方式已被确定为 COVID-19 期间心理健康的预测因素。³我们之前的研究证明了心理测量与大脑网络功能连接 (FC) 之间相互作用的相关性。⁴然而，重要问题仍有待解决。例如，FC（单独或与其他措施结合）可以预测长期心理健康吗？此外，大多数研究都集中在情感方面（心理困扰），尽管心理健康（MH）包括情感、心理（个人成长，[PG]）和社交（孤独）福祉组成部分，这些组成部分在大流行期间受到不同程度的影响。³本研究旨在调查 FC 测量值和 MH 组成部分的长期变化之间是否存在特异性，了解大脑网络和“弹性过程”之间的联系。^{5, 6}

我们研究了 702 名符合 Bayes-Marin 及其同事 2023 年研究标准的健康中年个体（350 名女性，年龄：50.66 ± 6.98 岁）。³所有参与者均根据赫尔辛基宣言签署了书面知情同意书。该研究方案得到了加泰罗尼亚医院联合基金会研究委员会 (CEIC-17/06) 的批准。对 COVID-19 爆发前采集的静息态功能磁共振成像图像进行了预处理，并计算了七个静息态网络 (RSN) 的系统隔离（SyS；集成-隔离平衡）[参见⁴ ] 和数据 S1）。拟合多项逻辑回归来预测 MH 三个组成部分的轨迹隶属关系（心理困扰和孤独的弹性或慢性轨迹，以及 PG 的弹性、逐渐上升或恶化，如通过生长混合模型捕获的，对比大流行前与大流行期间）两年随访期间的观察结果（参见 3，数据 S1和图 1a）。RSN 模型包括七个 RSN 的 FC。完整模型结合了显着的 RSN 测量值和我们之前研究中发现的显着预测因素（年龄、性别、月收入、压力应对、性格、一般健康状况和生活习惯）（参见³）。非 RSN 模型作为完整模型，但没有 RSN 数据，通过似然比检验（非 RSN 与完整模型），我们评估拟合优度是否良好通过将 FC 测量添加到社会人口统计学、心理和生活方式测量中来改进。总共，我们为 MH 的三个组成部分中的每一个组成部分安装了三个模型（RSN、非 RSN 和完整）。

情绪轨迹成员资格（图 1a）由显着网络（SN）FC（图 1b）显着预测，表明与慢性和慢性相比，功能更整合的 SN（即较低的 SyS）更能代表弹性轨迹。略高于中等弹性。在完整模型中，SN-FC 作为预测因子的重要性降低，并且完整模型与非 RSN 模型的比较不显着（图 1c）。

心理轨迹隶属度（图1a）由背侧注意网络（DAN）显着预测，而边缘网络（LN）和额顶网络（FPN）则稍稍预测。然而，在完整模型中，只有 FPN-FC 保持严格显着性，具有更高的概率属于具有更大集成 FPN 的弹性轨迹。值得注意的是，完整模型明显优于非 RSN 模型（图 1c）。

最后，任何 RSN-SyS 值都不能显着预测社会轨迹成员资格 ( R ²  = 0.008)。

我们的研究结果表明，反映主要 RSN 整合分离的 FC 指标为整个 COVID-19 大流行期间的长期 MH 结果提供了不同的预测。这项研究表明，大流行性焦虑抑郁趋势受到 SN-SyS 的影响。然而，情绪轨迹是通过一个更简单且信息量相同的模型预测的，该模型不包含 RSN 信息，这表明生活方式和心理因素足以描述它们。然而，在研究 PG 时，基线 FPN-SyS 被证明可以为从汇总的社会人口统计、心理因素和生活方式中得出的模型添加有意义的信息。 PG 维持的个体差异可能反映了通过重新评估和重视压力情况而蓬勃发展的能力。⁷因此，PG 和 FPN 连接之间发现的关联通常与认知灵活性和控制过程相关，⁸可能揭示了认知在心理方面的重要性。最后，没有发现任何与社会福祉相关的 RSN-SyS 可能是由于测量了个人的主观感知，而不是直接参与社会接触，和/或之前报道的疫情爆发期间疫情对孤独感的矛盾影响。⁹总而言之，目前的发现有助于我们之前的观察，⁴揭示了 DMN-和 FPN-SyS 通过压力感知对情绪轨迹的影响。这支持了三网观点。^5、6、10

总的来说，我们的研究结果表明，评估大脑网络整合与隔离有助于预测个体在 MH 维度上的弹性和脆弱性，从而能够及早识别高危个体，并设计和评估个性化预防策略。