Scholarlink

Background <p>Late-life depression (LLD) often co-occurs with mild cognitive impairment (MCI), and patients with LLD and MCI (LLD-MCI) have an increased risk of progression to Alzheimer’s disease (AD). However, differences in resting-state neural oscillation and cognitive impairment in LLD patients remain unclear. In this cross-sectional study, electroencephalography (EEG) was used to analyse, local rhythm activity and large-scale network communication to differentiate LLD patients with and without MCI.</p> Methods <p>We enrolled 113 participants: 74 with LLD (50 with LLD-MCI and 24 with LLD-non-MCI) and 39 healthy older adults (HOAs). All participants underwent comprehensive neuropsychological assessments. Spectral power and source-level functional connectivity (Phase-Locking Value, PLV) were analysed across multiple frequency bands. A machine learning framework using nested stratified cross-validation was implemented to evaluate the potential of EEG features in classifying LLD clinical subtypes.</p> Results <p>LLD-MCI patients exhibited a distinct dissociation in the beta band: significantly reduced spectral power in the left frontal cortex contrasted with extensive hyperconnectivity primarily centred on the right lateral orbitofrontal cortex (rLOFC). Complementary analyses also revealed widespread hyperconnectivity in the theta band in the LLD-MCI group. The Linear Discriminant Analysis (LDA) model achieved superior performance in distinguishing LLD-MCI patients from LLD-non-MCI patients, with an area under the curve (AUC) of 0.82 and an accuracy of 78.38%. Feature importance analysis revealed rLOFC-mediated beta synchronisation as the most discriminative biomarker.</p> Conclusion <p>Our findings suggest that beta-band oscillatory disruption—characterised by local power deficits and network hyperconnectivity—may represent a potential neurobiological signature of cognitive vulnerability in LLD patients. Whether this hyperconnectivity reflects a compensatory or pathological process remains a hypothesis for further validation. EEG metrics provide significant diagnostic value for the precise clinical subtyping and early identification of cognitive decline in the LLD population.</p> Clinical trial number <p>Not applicable.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

A multifeature machine learning and resting-state EEG study reveals differences in beta oscillation in late-life depression with or without mild cognitive impairment

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年5月7日

查看全文
  • Shihan Tian,
  • Gaohong Lin,
  • Yijie Zeng,
  • Qiang Wang,
  • Yicheng Lin,
  • Kexin Yao,
  • Zihan Wang,
  • Zhidai Xiao,
  • Jiayu Chen,
  • Youxuan Zheng,
  • Haoye Tan,
  • Zhangying Wu,
  • Mingfeng Yang,
  • Danyan Xu,
  • Xuanzi Li,
  • Jingyi Lao,
  • Shuang Liang,
  • Yunheng Chen,
  • Qin Liu,
  • Jiafu Li,
  • Yujing Gan,
  • Mingyong Zeng,
  • Ben Chen,
  • Yuping Ning,
  • Xiaomei Zhong

摘要

Background

Late-life depression (LLD) often co-occurs with mild cognitive impairment (MCI), and patients with LLD and MCI (LLD-MCI) have an increased risk of progression to Alzheimer’s disease (AD). However, differences in resting-state neural oscillation and cognitive impairment in LLD patients remain unclear. In this cross-sectional study, electroencephalography (EEG) was used to analyse, local rhythm activity and large-scale network communication to differentiate LLD patients with and without MCI.

Methods

We enrolled 113 participants: 74 with LLD (50 with LLD-MCI and 24 with LLD-non-MCI) and 39 healthy older adults (HOAs). All participants underwent comprehensive neuropsychological assessments. Spectral power and source-level functional connectivity (Phase-Locking Value, PLV) were analysed across multiple frequency bands. A machine learning framework using nested stratified cross-validation was implemented to evaluate the potential of EEG features in classifying LLD clinical subtypes.

Results

LLD-MCI patients exhibited a distinct dissociation in the beta band: significantly reduced spectral power in the left frontal cortex contrasted with extensive hyperconnectivity primarily centred on the right lateral orbitofrontal cortex (rLOFC). Complementary analyses also revealed widespread hyperconnectivity in the theta band in the LLD-MCI group. The Linear Discriminant Analysis (LDA) model achieved superior performance in distinguishing LLD-MCI patients from LLD-non-MCI patients, with an area under the curve (AUC) of 0.82 and an accuracy of 78.38%. Feature importance analysis revealed rLOFC-mediated beta synchronisation as the most discriminative biomarker.

Conclusion

Our findings suggest that beta-band oscillatory disruption—characterised by local power deficits and network hyperconnectivity—may represent a potential neurobiological signature of cognitive vulnerability in LLD patients. Whether this hyperconnectivity reflects a compensatory or pathological process remains a hypothesis for further validation. EEG metrics provide significant diagnostic value for the precise clinical subtyping and early identification of cognitive decline in the LLD population.

Clinical trial number

Not applicable.