Deconvolution of bolus-tracking data: a comparison of discretization methods

Area of Science:

• Medical Imaging
• Biophysics
• Pharmacokinetics

Background:

• Model-free perfusion measurement from bolus-tracking data necessitates tracer kinetic model discretization.
• Various discretization approaches exist, each with potential impacts on accuracy.
• Understanding these methods is crucial for reliable perfusion quantification in medical imaging.

Purpose of the Study:

• To classify and compare the accuracy of different tracer kinetic model discretization methods.
• To evaluate method performance under various physiological and imaging conditions, including different tissue models, delays, dispersion, temporal resolution, and signal-to-noise ratios.
• To provide guidance on selecting the optimal discretization method for specific clinical or research scenarios.

Main Methods:

• Classification of existing tracer kinetic model discretization approaches into delay-invariant (circulant, time shift) and non-delay-invariant (Volterra, singular, hybrid) categories.
• Simulations of magnetic resonance imaging (MRI) brain perfusion using plug flow and compartment tissue models with varied parameters.
• Comparison of simulation results with a patient dataset.

Main Results:

• Both delay-invariant methods (circulant and time shift) demonstrated comparable accuracy, though the circulant method showed sensitivity to data truncation.
• The Volterra method yielded the highest perfusion estimates, followed by hybrid, singular, and delay-invariant methods.
• Volterra's accuracy was superior except in unrealistic plug-flow scenarios without delay or dispersion; differences diminished with increased delay/dispersion relative to temporal resolution.

Conclusions:

• The time shift method is recommended when negative delays are unavoidable or precise left-right perfusion ratios are critical.
• For positive delays and a focus on absolute accuracy, the Volterra method is preferred.
• Method selection depends on specific physiological conditions and the desired quantitative outcome in perfusion imaging.