Prediction of health anxiety using resting-state functional near-infrared spectroscopy and machine learning

The role of cortical networks in health anxiety remain poorly understood. This study aimed to develop a predictive model for health anxiety, using a machine-learning approach based on resting-state functional connectivity (rsFC) with functional near-infrared spectroscopy (fNIRS).

One hundred and four university students experiencing school disclosure due to the Corona Virus Disease 2019 pandemic participated in the study, and the final sample consisted of 90 participants. All participants underwent a 6-min resting-state fNIRS recording session and filled out the Short Health Anxiety Inventory after the data collection. Stratified 10-fold cross-validation was used to train and evaluate the Lasso regression model. Additionally, a bootstrap method was used to determine which features significantly contributed to the prediction of health anxiety.

The contributing rsFC with negative weights was the functional connectivity between right medial superior frontal gyrus and right middle frontal gyrus, with a 99 % confidence interval (CI) of [-1.61, -0.35]. The contributing rsFC with positive weights was the functional connectivity between right supramarginal gyrus and left middle temporal gyrus (99 % CI = [0.02, 1.67]), as well as the functional connectivity between right medial superior frontal gyrus and right supramarginal gyrus (99 % CI = [0.03, 1.41]).

The findings reveal a predictive role of intrinsic cortical organization in health anxiety and suggest that health anxiety involves complex interactions between cognitive control, emotion regulation, and sensory processing. The work provides new insights into potential neural mechanisms underlying health anxiety, and implications for neuromodulation research and practice targeting severe health anxiety.