Reconstruction-Aware Imaging System Ranking by Use of a Sparsity-Driven Numerical Observer Enabled by Variational...

Area of Science:

• Medical Imaging
• Image Quality Assessment
• Signal Processing

Background:

• Task-based image quality measures guide imaging system optimization.
• Traditional ideal observers optimize hardware independently of reconstruction methods.
• Accurate object statistics models are often challenging to determine.

Purpose of the Study:

• To introduce and investigate a sparsity-driven observer (SDO) for optimizing imaging hardware.
• To develop a method where hardware optimization and sparse image reconstruction are intrinsically linked.
• To utilize a stochastic object model that captures object sparsity.

Main Methods:

• Developed a sparsity-driven observer (SDO) utilizing a stochastic object model.
• Employed variational Bayesian inference tools for efficient computation of the SDO test statistic.
• Applied the SDO to rank data-acquisition designs in a magnetic resonance imaging-inspired scenario.

Main Results:

• The SDO effectively optimizes hardware by matching it with sparse reconstruction techniques.
• SDO utilizes shared statistical information about object sparsity for both observer and reconstruction.
• SDO-derived rankings align with visual assessments of reconstructed images.

Conclusions:

• The SDO offers a unified approach to optimize imaging hardware and reconstruction in compressive sensing systems.
• SDO provides an alternative to traditional observers like the Hotelling observer, yielding different, potentially more relevant, rankings.
• This approach enhances task-based image quality optimization by integrating object sparsity information.