Personalized Models for Injected Activity Levels in SPECT Myocardial Perfusion Imaging

Area of Science:

• Nuclear Medicine
• Radiology
• Medical Imaging

Background:

• Myocardial perfusion imaging (MPI) using SPECT is crucial for diagnosing heart conditions.
• Current MPI protocols use a standard radiotracer dose, potentially leading to unnecessary radiation exposure in some patients.
• Optimizing injected activity is essential for balancing image quality and radiation safety.

Purpose of the Study:

• To develop and validate a patient-specific approach for reducing injected radiotracer activity in SPECT-MPI.
• To utilize machine learning to predict individualized minimum dose levels based on patient characteristics.
• To assess the impact of personalized dosing on image quality and diagnostic accuracy for different reconstruction methods.

Main Methods:

• A dataset of 857 clinical SPECT-MPI acquisitions was used to establish minimum dose levels for image quality preservation.
• Machine learning models were trained using patient attributes (body measurements, clinical variables) to predict these personalized dose levels.
• The personalized dose strategy was evaluated for Filtered Backprojection (FBP) and Ordered-Subsets Expectation-Maximization (OS-EM) reconstruction methods.
• Perfusion defect detection accuracy was assessed using the Quantitative Perfusion SPECT software.

Main Results:

• Individualized dose reductions varied significantly compared to conventional clinical doses.
• Personalized dose models achieved greater average dose reduction than a non-patient-specific approach.
• For FBP and OS-EM, average personalized doses were reduced to 58% and 54% of the full clinical dose, respectively, maintaining diagnostic accuracy.
• OS-EM with personalized dosing at 16% of full dose achieved accuracy comparable to FBP at full dose.

Conclusions:

• Patient-specific tailoring of injected activity in SPECT-MPI is feasible and effective for dose reduction.
• Machine learning models can accurately predict personalized radiotracer doses based on patient data.
• Personalized dosing strategies, particularly with advanced reconstruction like OS-EM, offer substantial radiation dose savings while preserving diagnostic performance.