Back

Deep Learning for Predicting Cognitive Gap as a Reliable Biomarker of Dementia

Habuza, T.; Zaki, N.; Statsenko, Y.; Alnajjar, F.; Elyassami, S.

2021-01-26 radiology and imaging
10.1101/2021.01.24.21249625 medRxiv
Show abstract

Neuroimaging data may reflect the mental status of both cognitively preserved individuals and patients with neurodegenerative diseases. To find the relationship between cognitive performance and the difference between predicted and observed functional test results, we developed a Convolutional Neural Network (CNN) based regression model to estimate the level of cognitive decline from preprocessed T1-weighted MRI images. In this study, we considered the Predicted Cognitive Gap (PCG) as the biomarker to accurately classify Healthy Control (HC) subjects versus Alzheimer disease (AD) subjects. The proposed model was tested on a dataset that includes 422 HC and 377 AD cases. The performance of the proposed solution was measured using Receiver Operating Characteristic (ROC) Area Under the Curve (AUC) and achieved 0.987 (ADAS-cog), 0.978 (MMSE), 0.898 (RAVLT), 0.848 (TMT), 0.829 (DSST) for averaged brain images; and 0.985 (ADAS-cog), 0.987 (MMSE), 0.901 (RAVLT), 0.8474 (TMT), 0.796 (DSST) for middle slice skull stripped brain images. The results achieved indicate that PCG can accurately separate healthy subjects from demented ones and thus, the structure of the brain contributes to the level of human cognition and their functional abilities. Therefore, PCG could be used as a biomarker for dementia.

Matching journals

The top 5 journals account for 50% of the predicted probability mass.

50% of probability mass above

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.