False Discovery Rate Control for Lesion-Symptom Mapping With Heterogeneous Data via Weighted p-Values

Area of Science:

• Neuroscience
• Cognitive Science
• Biostatistics

Background:

• Lesion-symptom mapping investigates brain regions' roles in cognition using patient data with natural brain injuries.
• Observational data in these studies often exhibit nonuniform lesion distribution, leading to power heterogeneity in statistical tests.
• Traditional mass univariate tests lack power in regions with sparse lesion data.

Purpose of the Study:

• To propose and evaluate a p-value weighting method for voxel-based lesion-symptom mapping (VLSM).
• To address power heterogeneity and improve statistical power in VLSM studies.
• To enhance the identification of brain regions associated with cognitive deficits.

Main Methods:

• Developed a p-value weighting approach using lesion distribution and spatial information.
• Estimated non-null prior probabilities for hypothesis tests based on lesion data.
• Introduced a monotone minimum weight criterion requiring minimal a priori power information.
• Validated methods using simulated data and an aphasia study in stroke survivors.

Main Results:

• The proposed weighting method demonstrated robust error control.
• The approach successfully increased statistical power in VLSM analyses.
• Effectively identified brain regions with inconclusive results due to low power.
• Showcased improved identification of lesion-symptom associations.

Conclusions:

• P-value weighting is a powerful tool for enhancing VLSM studies.
• The method improves statistical power and error control in lesion-brain mapping.
• This approach offers a way to better understand brain-behavior relationships by leveraging lesion data more effectively.