Unscaled Indices for Assessing Agreement of Functional Data

Area of Science:

• Biostatistics
• Medical Imaging Analysis
• Health Technology Assessment

Background:

• Adopting new medical devices necessitates rigorous validation of clinical measurement reliability and reproducibility.
• Functional data generated by high-tech medical devices require robust methods for assessing agreement across different measurement techniques.

Purpose of the Study:

• To establish validity and acceptability of modern high-tech medical devices generating functional data.
• To introduce novel indices for assessing agreement of multiple functional data measured by different methods/technologies/raters on the same subjects.
• To develop methods for delineating intramethod, intermethod, and total agreement trends over time.

Main Methods:

• Introduction of unscaled indices: total deviation index (TDI) and coverage probability (CP), which are functions of time.
• Development of scalar-valued TDI and CP indices summarizing agreement over the entire domain using weighted averages.
• Utilizing a bivariate multilevel functional linear mixed effects model to express indices based on mean functions and variance components.
• Employing bivariate multilevel functional principal component analysis for smooth index estimation via eigenanalyses of univariate covariance functions.

Main Results:

• The proposed indices (TDI and CP) can delineate trends of intramethod, intermethod, and total agreement across time on the original measurement scale.
• Scalar-valued indices provide a summary of agreement over the entire data domain.
• The functional principal component analysis approach ensures efficient and scalable estimation.
• Simulation studies confirmed the finite-sample properties of the proposed estimators.

Conclusions:

• The developed indices and methods provide a rigorous framework for assessing the reliability and reproducibility of functional data from medical devices.
• The approach is applicable to diverse high-tech medical imaging devices, such as those used in diuresis renography for kidney obstruction detection.
• This work contributes to the evidence-based adoption of advanced medical technologies by ensuring measurement consistency and validity.