Dynamic functional connectivity changes in noise-induced hearing loss: A resting-state fMRI study with machine learning-based classification.

OBJECTIVE

Noise-induced hearing loss impacts brain health and cognition, with dynamic functional connectivity analysis offering a promising but underexplored method for studying whole-brain activity. Therefore, this study aimed to utilise dynamic functional connectivity analysis to investigate abnormal temporal variability in whole-brain functional connectivity in patients with noise-induced hearing loss.

METHODS

In this observational study, 58 patients with noise-induced hearing loss and 42 healthy male controls, matched for age and education, underwent resting-state functional magnetic resonance imaging. The sliding window approach was employed to evaluate dynamic functional connectivity between region pairs, and k-means clustering was used to identify dynamic functional connectivity states. A two-sample t-test was used to compare differences in dynamic functional connectivity variability and state metrics between patients with noise-induced hearing loss and healthy male controls (P < 0.05). Abnormal brain dynamic functional connectivity features were identified using false discovery rate correction and least absolute shrinkage and selection operator classifier. These features were used to construct support vector machine classifiers.

RESULTS

Compared with healthy male controls, patients with noise-induced hearing loss demonstrated decreased dynamic functional connectivity between the right supplementary motor area and bilateral cuneus and increased dynamic functional connectivity between the supplementary motor area and left inferior parietal gyrus. The support vector machine classifier based on abnormal dynamic functional connectivity features selected by false discovery rate correction successfully distinguished between patients with noise-induced hearing loss and healthy male controls with an accuracy of 82.5%. The accuracy of the support vector machine classifier based on least absolute shrinkage and selection operator-selected abnormal dynamic functional connectivity features reached 96.8%.

CONCLUSION

This study revealed abnormal dynamic functional connectivity patterns in patients with noise-induced hearing loss, offering insights into the complex neuropathological mechanisms underlying long-term brain network changes associated with this disease.