Classification of mild cognitive impairment using convolutional neural network based on functional near-infrared spectroscopy-derived neuroimaging biomarkers

Abstract

Abstract Background To date, early detection of mild cognitive impairment (MCI) has mainly depended on paper-based neuropsychological assessments. Recently, biomarkers for MCI detection has gained a lot of attention because of the low sensitivity of neuropsychological assessments. This study proposed the functional near-infrared spectroscopy (fNIRS)-derived neuroimaging technique to identify MCI using convolutional neural network (CNN). Methods Eighty subjects with MCI and 142 healthy controls (HC) performed the 2-back task, and their oxygenated hemoglobin (HbO2) changes in the dorsolateral prefrontal cortex (DLPFC) were recorded during the task. CNN was applied to distinguish MCI from HC after training the CNN model with spatial features of brain images within the time window during 5–15 seconds. Thereafter, the 5-fold cross-validation approach then was used to evaluate the performance of CNN. Results Significant difference in averaged HbO2 values between MCI and HC groups were found, and the average accuracy of CNN was 95.71%. Specifically, the left DLPFC (98.62%) achieved a higher accuracy rate than the right DLPFC (92.86%). Conclusion These findings suggest that the fNIRS-derived neuroimaging technique based on the spatial feature could be a promising way for early detection of MCI.