Noise Reduction in Brain CT: A Comparative Study of Deep Learning and Hybrid Iterative Reconstruction Using Multiple...

Area of Science:

• Medical Imaging
• Radiology
• Image Processing

Background:

• Computed tomography (CT) is essential for brain imaging.
• Noise in CT images can degrade diagnostic quality.
• Advanced reconstruction techniques aim to improve image quality and reduce noise.

Purpose of the Study:

• To compare the noise reduction capabilities of deep learning reconstruction (DLR) and hybrid iterative reconstruction (HIR) in brain CT.
• To evaluate the influence of parameters like slice thickness and tube current on noise reduction for both methods.

Main Methods:

• CT images of phantoms and 11 patients were reconstructed using filtered backprojection (FBP), DLR, and HIR at various slice thicknesses (0.625-5 mm).
• Noise reduction ratio was quantified using FBP as a reference.
• Visual assessment of image quality compared DLR and HIR for images with similar noise reduction.

Main Results:

• Both DLR and HIR demonstrated increased noise reduction with higher reconstruction levels.
• DLR's noise reduction was significantly dependent on slice thickness, being more effective with thinner slices.
• HIR showed less dependence on slice thickness, and its performance varied less across different imaging objects compared to DLR.

Conclusions:

• The noise reduction efficacy of DLR in brain CT is influenced by slice thickness, tube current, and the imaging object.
• DLR shows particular promise for thin-slice brain CT imaging, potentially offering superior image quality.
• Clinical application of DLR requires consideration of these influencing factors for optimal results.