A detection model for cognitive dysfunction based on volatile organic compounds from a large Chinese community cohort

Abstract

Introduction: We explored whether volatile organic compound (VOC) detection can serve as a screening tool to distinguish cognitive dysfunction (CD) from cognitively normal (CN) individuals. Methods: The cognitive function of 1467 participants was assessed and their VOCs were detected. Six machine learning algorithms were conducted and the performance was determined. The plasma neurofilament light chain (NfL) was measured. Results: Distinguished VOC patterns existed between CD and CN groups. The CD detection model showed good accuracy with an area under the receiver-operating characteristic curve (AUC) of 0.876. In addition, we found that 10 VOC ions showed significant differences between CD and CN individuals (p < 0.05); three VOCs were significantly related to plasma NfL (p < 0.005). Moreover, a combination of VOCs with NfL showed the best discriminating power (AUC = 0.877). Discussion: Detection of VOCs from exhaled breath samples has the potential to provide a novel solution for the dilemma of CD screening.