Task-specific spatial resolution properties of iterative and deep learning-based reconstructions in computed...

Area of Science:

• Medical Imaging
• Radiology
• Image Reconstruction

Background:

• Computed tomography (CT) image reconstruction techniques like iterative reconstruction (IR) and deep-learning-based image reconstruction (DLIR) aim to improve image quality.
• Evaluating the true transfer function (TTF) is crucial for assessing the performance of these reconstruction methods, especially for small structures like enhanced vessels.

Purpose of the Study:

• To evaluate the TTFs of DLIR and IR compared to conventional filtered back projection (FBP) in CT.
• To introduce and validate a novel method for assessing TTFs of small simulated enhanced vessels (1 mm diameter).

Main Methods:

• A 1 mm iodine wire phantom was used to simulate small enhanced vessels, and its TTF was measured using a Fourier transform-based iodine wire (IW) method.
• Conventional methods, circular edge (CE) and plane edge (PE), were used for a 30 mm rod phantom.
• TTFs were evaluated for DLIR, IR, and FBP on 256-row and 64-row CT systems at different dose levels.

Main Results:

• The IW method revealed significantly lower TTFs for DLIR and IR compared to FBP, indicating blurring of small vessels.
• TTFs measured by the IW method were notably lower than those obtained using CE or PE methods.
• Both DLIR and IR demonstrated blurring of 1 mm diameter simulated vessels.

Conclusions:

• The proposed IW method is effective for evaluating TTFs of small enhanced vessels in CT, overcoming limitations of conventional CE and PE methods.
• Conventional TTF evaluation methods may not accurately represent the performance of modern CT reconstruction algorithms for small vessel imaging.