CNN-based PET sinogram repair to mitigate defective block detectors

Area of Science:

• Medical Imaging
• Artificial Intelligence in Healthcare
• Nuclear Medicine

Background:

• Positron emission tomography (PET) scanners utilize block detectors to enhance sensitivity and axial coverage.
• Aging or malfunctioning block detectors can lead to data loss, resulting in artifacts and image quality degradation.
• Existing methods for addressing data loss in PET imaging are limited.

Purpose of the Study:

• To propose and evaluate a deep convolutional neural network (CNN) for sinogram repair in PET imaging.
• To mitigate the impact of malfunctioning block detectors on PET image quality and quantitative accuracy.
• To demonstrate the superiority of the proposed CNN method over existing techniques.

Main Methods:

• Development of a deep convolutional neural network (CNN) for sinogram repair.
• Experimental validation using whole-body patient PET studies with simulated data loss.
• Comparison of the CNN method against previously published techniques using quantitative metrics.

Main Results:

• The proposed CNN method significantly outperforms existing methods in sinogram repair.
• Normalized mean squared error for raw sinograms is substantially reduced.
• Multi-scale structural similarity and quantitative accuracy of reconstructed PET images are significantly improved.

Conclusions:

• Deep convolutional neural networks offer a powerful solution for repairing missing sinogram data caused by malfunctioning PET detectors.
• The proposed CNN-based approach effectively restores image quality and quantitative accuracy in PET scans.
• This method holds significant potential for improving the reliability and diagnostic value of PET imaging.