Voxel-based multimodel fitting method for modeling time activity curves in SPECT images.

Area of Science:

• Nuclear Medicine
• Medical Imaging Analysis
• Pharmacokinetics

Background:

• Accurate biodistribution and pharmacokinetic estimation from SPECT images is crucial for dose calculation and understanding activity distribution.
• Current methods are limited by region-based uniformity assumptions or single-model voxel-based approaches.

Purpose of the Study:

• To develop and validate a voxel-based multimodel fitting method (VoMM) for precise per-voxel pharmacokinetic analysis from time-sequence SPECT data.
• To compare VoMM's performance against traditional region-based and single-model voxel-based methods.

Main Methods:

• Proposed a voxel-based multimodel fitting method (VoMM) that selects the optimal fitting model for each voxel using Akaike criteria.
• Introduced a restricted method based on trapezoid integration to manage noisy data and assess fitting failures.
• Validated the method through numerical experiments and real patient SPECT data.

Main Results:

• VoMM achieved comparable root mean square errors to bi-exponential fitting and significantly lower errors than single-model approaches (<5% vs. >10%).
• VoMM demonstrated lower failure rates in fitting, successfully analyzing 90% of voxels in clinical data.
• Organ-based time-integrated activity (TIA) estimations were similar to ROI-based methods, but mean residence time calculations showed larger differences (up to 25%).

Conclusions:

• The voxel-based multimodel fitting method (VoMM) is a feasible and robust approach for analyzing patient SPECT data.
• VoMM provides accurate organ-based TIA estimations and is advantageous for pharmacokinetic analysis, spatial heterogeneity studies, and voxel-based absorbed dose assessments by preventing model overfitting.