Receiver operating characteristic (ROC) analysis of images reconstructed with iterative expectation maximization...

Area of Science:

• Medical Imaging
• Nuclear Medicine

Background:

• Image reconstruction is crucial for single-photon emission computed tomography (SPECT).
• Maximum Likelihood-Expectation Maximization (ML-EM) and Ordered Subset (OS)-EM are iterative algorithms used for SPECT image reconstruction.
• Filtered Back Projection (FBP) is a traditional method for SPECT image reconstruction.

Purpose of the Study:

• To evaluate and compare the image quality of ML-EM and OS-EM algorithms for SPECT.
• To determine optimal parameters (iterations, subsets) for OS-EM algorithm.
• To compare OS-EM and ML-EM with FBP for SPECT image reconstruction.

Main Methods:

• Utilized phantoms mimicking SPECT imaging of cerebral blood flow, myocardial perfusion, bone, and tumor.
• Employed Receiver Operating Characteristic (ROC) analysis to evaluate image recognition with varying iterations and subsets.
• Applied ROC analysis results to myocardial phantom experiments and perfusion scintigraphy.

Main Results:

• OS-EM algorithm with 10 iterations and 5 subsets yielded good SPECT image quality.
• This combination balanced image quality with acceptable image processing time.
• ROC analysis guided the selection of optimal parameters for OS-EM.

Conclusions:

• The study provides guidance on selecting optimal parameters (iterations and subsets) for ML-EM and OS-EM algorithms in SPECT imaging.
• OS-EM demonstrates a favorable balance between image quality and processing efficiency.
• Findings aid in optimizing SPECT image reconstruction for improved diagnostic accuracy.