Scanning linear estimation: improvements over region of interest (ROI) methods

Area of Science:

• Medical Imaging
• Tomography
• Signal Processing

Background:

• Tomographic medical imaging typically estimates signal activity by summing voxels from reconstructed images.
• This conventional approach can be limited by noise, background randomness, and system variability.

Purpose of the Study:

• To introduce and evaluate an alternative signal activity estimation scheme operating on raw projection data.
• To demonstrate substantial improvements in ensemble mean-square error (EMSE) compared to existing methods.

Main Methods:

• Developed a scanning-linear (SL) estimator as a tractable approximation of maximum-likelihood estimation.
• The SL estimator accounts for background randomness, measurement noise, and parameter variability.
• Simulated noisy projection data using calibration data from a small-animal SPECT imaging system.

Main Results:

• The SL estimator demonstrated unbiased signal activity estimation when signal size and location are known.
• Compared to maximum-likelihood expectation-maximization (MLEM) reconstructions, the SL method showed dramatic improvements in EMSE for both signal estimation tasks.
• The SL method proved effective for simultaneously estimating signal size, location, and activity, as well as for estimating activity with fixed signal parameters.

Conclusions:

• The scanning-linear (SL) estimator offers a significant advancement in tomographic medical imaging by utilizing raw projection data.
• This novel approach substantially reduces estimation errors and bias compared to conventional methods relying on reconstructed images.
• The SL method provides a more accurate and reliable tool for quantifying signal activity, crucial for applications like monitoring therapeutic responses.