Identifying HIV Associated Neurocognitive Disorder Using Large-Scale Granger Causality Analysis on Resting-State...

Area of Science:

• Neuroscience
• Medical Imaging
• Machine Learning

Background:

• HIV infection can cause cognitive impairment (HAND).
• Current diagnosis relies on neuropsychological testing.
• Resting-state fMRI offers a non-invasive alternative to assess brain connectivity.

Purpose of the Study:

• To apply large-scale Granger Causality (lsGC) for multivariate brain information flow analysis.
• To evaluate lsGC's effectiveness in predicting HIV Associated Neurocognitive Disorder (HAND).
• To compare lsGC with conventional Granger causality for diagnostic accuracy.

Main Methods:

• Extracted pairwise multivariate information flow using lsGC from resting-state fMRI data.
• Utilized Generalized Matrix Learning Vector Quantization (GMLVQ) for classification.
• Compared lsGC-GMLVQ with conventional Granger causality for predicting HAND.

Main Results:

• The lsGC-GMLVQ approach achieved 87% accuracy and an AUC of 0.90.
• This represents a statistically significant improvement over conventional Granger causality (76% accuracy, 0.74 AUC).
• The multivariate method better captures altered brain interaction patterns in HAND.

Conclusions:

• lsGC provides a robust measure of brain connectivity for HAND prediction.
• The multivariate approach significantly outperforms conventional methods.
• This technique offers a promising non-invasive tool for diagnosing HAND-related cognitive impairment.