Background correction in factor analysis of dynamic scintigraphic studies: necessity and implementation

Area of Science:

• Data analysis
• Signal processing
• Scientific modeling

Background:

• Factor analysis separates structures based on temporal behavior, but background noise can obscure true signals.
• The region of interest (ROI) method uses background subtraction to address signal overlap.
• Factor analysis has historically overlooked the issue of total structure overlap.

Purpose of the Study:

• To demonstrate the inaccuracies of standard factor analysis when dealing with total structure overlap.
• To introduce an improved factor analysis method that accounts for overlapping structures.
• To compare the novel method with the classical ROI approach.

Main Methods:

• Developed a modified factor analysis approach by assuming local homogeneity within overlapping structures.
• Applied the improved method to datasets with significant background overlap.
• Compared the results against the traditional ROI method for accuracy and specificity.

Main Results:

• Proved that standard factor analysis produces erroneous results with total structure overlap.
• The proposed method, assuming local homogeneity, significantly enhances the accuracy of the factor analysis solution.
• The new approach is operator independent, unlike the ROI method.

Conclusions:

• The assumption of local homogeneity is crucial for accurate factor analysis in the presence of overlapping signals.
• The developed method provides a more robust and specific alternative to the ROI technique for analyzing complex data.
• This advancement has implications for various fields relying on factor analysis for signal separation and interpretation.