An fNIRS representation and fNIRS-scales multimodal fusion method for auxiliary diagnosis of amnestic mild cognitive impairment

Amnestic mild cognitive impairment (aMCI) is the prodromal period of more serious neurodegenerative diseases (e.g., Alzheimer's disease), characterized by declines in memory and thinking abilities. Auxiliary assessment and early diagnosis of aMCI are crucial in preventing the continued deterioration of cognitive abilities; nevertheless, this task poses a formidable challenge due to the inconspicuous nature of early symptoms. Functional near-infrared spectroscopy (fNIRS) is a non-invasive, low-cost, and user-friendly neuroimaging technique, which is capable of detecting subtle changes in brain activity among different subjects. Moreover, multimodal fusion can assess cognition status from different perspectives and enhance auxiliary diagnosis accuracy significantly. This paper proposes an fNIRS representation and fNIRS-scales multimodal fusion method for auxiliary diagnosis of aMCI. Specifically, we convert one-dimensional time-series fNIRS signals into two-dimensional images with Gramian Angular Field and achieve end-to-end fNIRS representation with convolutional neural network. Then, we integrate the extracted features with cognitive scales at the decision-making level to improve the diagnosis accuracy of aMCI, employing the data balance strategy to prevent biased prediction. What is more, based on the fNIRS features, we also propose a data-driven cognitive scales-screening method to help the physician to assess aMCI with higher efficiency. We conducted experiments on 86 subjects (including 53 aMCI patients and 33 normal controls) recruited from Foshan First People's Hospital. The diagnosis accuracy reaches 88.02% and 93.90% with fNIRS representation and further fNIRS-scales fusion, respectively. With the cognitive scales-screening, we delete 50% scales, reducing test time but only losing 2.54% accuracy.