Optimal likelihood-ratio multiple testing with application to Alzheimer's disease and questionable dementia

Area of Science:

• Neuroimaging
• Statistical Methods
• Medical Diagnostics

Background:

• Controlling the false discovery rate (FDR) is crucial for multiple hypothesis testing in scientific research.
• Enhancing the detection power of FDR control methods is essential for robust scientific discovery.

Purpose of the Study:

• To introduce and evaluate a novel likelihood ratio-based multiple testing method for neuroimage analysis.
• To compare the performance of this new method against conventional techniques, specifically the Benjamini-Hochberg FDR method.

Main Methods:

• The likelihood ratio-based FDR method was assessed using simulated data under an independence assumption.
• Performance was evaluated on positron emission tomography (PET) data from patients with Alzheimer's disease and questionable dementia.
• The method's ability to detect extensive hypometabolic regions was compared to the Benjamini-Hochberg FDR method.

Main Results:

• The likelihood ratio-based FDR method demonstrated effective control of the false discovery rate.
• It yielded the smallest false non-discovery rate for one-sided tests and the smallest expected number of false assignments for two-sided tests.
• In Alzheimer's disease patients, the method detected more extensive hypometabolic regions than the Benjamini-Hochberg method, consistent with previous findings.
• Hypometabolism in the medial temporal region was identified in questionable dementia patients using the proposed method.

Conclusions:

• The likelihood ratio-based FDR method offers efficient identification of extensive hypometabolic regions in neuroimaging studies.
• Its enhanced detection capability and robust FDR control make it a valuable tool for analyzing complex neurobiological data.