[A non-local means approach for PET image denoising]

Area of Science:

• Medical image processing
• Radiological imaging analysis
• Computational imaging

Context:

• Medical image denoising is critical for accurate diagnosis.
• Positron Emission Tomography (PET) imaging is susceptible to noise, impacting image quality and diagnostic interpretation.
• Existing denoising methods like median and Wiener filtering have limitations in preserving fine details.

Purpose:

• To adapt and evaluate the Non-local means algorithm for denoising clinical PET images.
• To compare the performance of the Non-local means method against traditional filtering techniques (median and Wiener filtering).
• To assess the algorithm's ability to suppress noise while maintaining structural integrity for diagnostic purposes.

Summary:

• The Non-local means algorithm was adapted from image processing principles for PET image denoising.
• Experimental results on real clinical PET data demonstrated superior performance compared to median and Wiener filtering.
• The proposed method effectively suppresses noise and preserves diagnostically important structural information in PET images.

Impact:

• Provides an effective denoising solution for PET imaging, enhancing diagnostic accuracy.
• Offers a valuable tool for radiologists and researchers working with PET data.
• Contributes to the advancement of medical image processing techniques for nuclear medicine.