MENU

Click here to

×

Are you sure ?

Yes, do it No, cancel

A Deep Learning Model to Predict a Diagnosis of MCI by Using Static and Dynamic Brain Connectomics

D Cui1,2, J Jin2, Z Liu2, T Yin2*, (1) Shandong First Medical University, Taian, 37, CN, (2) Institute Of Biomedical Engineering, Chinese Academy Of Medical Sciences, Tianjin, CN

Presentations

(Sunday, 7/12/2020)   [Eastern Time (GMT-4)]

Room: AAPM ePoster Library

Purpose: Mild cognitive impairment (MCI) is a transitional state between normal aging and dementia. Despite the importance of early diagnosis of Alzheimer's disease (AD) for prognosis and personalised interventions, we still lack robust tools for differentiating the borderline MCI and normal aging. Here, we propose a hybrid feature selection algorithm and constructed a multivariate pattern analysis (MVPA) auxiliary diagnosis model for MCI.
Methods: We used a cohort of 80 (36 females) early MCI subjects and 86 (41 females) normal subjects from the Alzheimer's disease Neuroimaging Initiative (ADNI) database. Based on a full data-driven approach, functional images were preprocessed using the DPABI and SPM12 toolkits. The stationary and dynamic functional connection network of 166 subjects were established using independent component analysis (ICA), and features were extracted. Then, a hybrid feature selection algorithm based on Kendall rank correlation coefficient method and Fisher criterion and linear discriminant classifier (LDC) were used to classify subjects with MCI.
Results: The misclassification rate decreased from 42.9% (Kendall sort correlation coefficient method) to 27.6% (hybrid feature selection algorithm) (P < 0.001), which indicates that the proposed hybrid feature selection algorithm can effectively improve the classification accuracy of MVPA. Further analysis of the steady-state experimental results demonstrates that features with stronger discrimination are mainly located in the prefrontal network, and default mode network. And the dynamic results revealed that the classification results were relatively good when the window width is large, and relatively stable when the step size s=5.
Conclusion: The proposed computer-aided diagnosis method highlights the potential of combining static and dynamic brain connectomics and deep learning to support clinical decision making in distinction of MCI vs. normal aging.

Funding Support, Disclosures, and Conflict of Interest: We are grateful for support from the Fundamental Research Funds for the Central Universities (3332018159), and the CAMS Initiative for Innovative Medicine (2016-I2M-1004).

Keywords

Brain, Dynamic Wedge, MRI

Taxonomy

TH- Dataset Analysis/Biomathematics: Machine learning techniques

Contact Email