A global estimator unbiased by local changes

Area of Science:

• Neuroimaging
• Biostatistics
• Medical Physics

Background:

• Global activity is crucial for statistical analysis in Positron Emission Tomography (PET) imaging, impacting error variance and sensitivity.
• Defining global activity as a simple average over all voxels introduces bias, leading to underestimation of true activations and artefactual deactivations in PET data.

Purpose of the Study:

• To develop a novel, unbiased estimator for global activity in PET data analysis.
• To improve the accuracy and reliability of functional imaging results by mitigating statistical biases inherent in traditional global activity measures.

Main Methods:

• Proposed a new estimator for global activity using singular value decomposition (SVD) to identify and rotate data modes.
• Employed a modified stochastic sign change (SSC) criterion to maximize data non-locality while preserving locality in remaining modes.
• Evaluated the estimator on simulated and real functional imaging (PET) data.

Main Results:

• The novel estimator significantly reduced bias introduced by focal activations by 80-90% compared to the global mean.
• Performance was comparable to a previously established unbiased estimator when tested on empirical PET data.
• The new method is independent of experimental design, relying on general signal characteristics.

Conclusions:

• The proposed SVD-based, SSC-optimized global activity estimator offers a more accurate and robust approach for PET data analysis.
• This method effectively addresses biases associated with traditional global activity measures, enhancing the validity of neuroimaging findings.
• The estimator's independence from experimental design makes it broadly applicable across various functional imaging studies.