A multimodal imaging-based integrative framework for HIV-associated cognitive impairment and treatment response

Area of Science:

• Neuroimaging
• HIV/AIDS Research
• Cognitive Science

Background:

• Combination antiretroviral therapy (cART) can improve cognition in people with HIV (PWH) by reducing inflammation.
• Cognitive improvements in PWH on cART are variable and not always evident in short-term studies.

Purpose of the Study:

• To develop and validate a multimodal integrative model to predict cognitive changes in PWH over 12 weeks of cART.
• To associate baseline neuroimaging metrics with neurocognitive performance and its longitudinal changes.

Main Methods:

• Applied a multimodal integrative model incorporating volumetric, cerebral blood flow, cerebrovascular reactivity, and diffusion MRI data.
• Utilized multilayered principal component analysis, penalized regression, and feature weight back-propagation for "large p, small n" data.
• Compared multimodal model performance against unimodal models for predicting cognitive change.

Main Results:

• Baseline imaging metrics showed modest associations with baseline cognitive scores, primarily driven by subcortical regions.
• Baseline imaging features were more strongly associated with longitudinal cognitive changes over 12 weeks of cART in PWH.
• The multimodal integrative model significantly outperformed unimodal models in explaining cognitive trajectories.

Conclusions:

• The developed interpretable statistical framework effectively integrates multimodal neuroimaging data.
• This approach offers a robust method for understanding short-term cognitive trajectories in PWH undergoing cART.
• Key brain regions like the frontal pole, amygdala, putamen, hippocampus, and posterior limb of the internal capsule were identified as important predictors.