Accelerated SPECT image reconstruction with FBP and an image enhancement convolutional neural network

Area of Science:

• Medical Imaging
• Nuclear Medicine
• Artificial Intelligence in Healthcare

Background:

• Monte Carlo reconstruction offers superior SPECT/CT image quality but is time-prohibitive for clinical use.
• Analytical correction methods are faster but less accurate than Monte Carlo techniques.
• Long reconstruction times limit the clinical adoption of advanced SPECT/CT imaging techniques.

Purpose of the Study:

• To develop a convolutional neural network (CNN) to accelerate image reconstruction in SPECT/CT.
• To improve the image quality of fast filtered back projection (FBP) reconstructions using AI.
• To achieve Monte Carlo-level image quality within clinically relevant timeframes.

Main Methods:

• A CNN was trained to enhance FBP images from pre-treatment SPECT/CT scans.
• Reconstruction methods compared included FBP, clinical reconstruction, Monte Carlo, and the CNN approach.
• Quantitative accuracy was assessed using phantom experiments and clinical data for hepatic radioembolization.

Main Results:

• The CNN achieved reconstructions in 5 seconds, significantly faster than clinical (5 min) and Monte Carlo (19 min) methods.
• CNN reconstruction quality, measured by mean squared error, fell between Monte Carlo and clinical methods.
• Lung shunting fraction differences were below 2 percent points, and quantitative measures for radioembolization were accurately retrieved.

Conclusions:

• Filtered back projection (FBP) combined with an image enhancement CNN yields high-quality SPECT reconstructions.
• The CNN approach provides near Monte Carlo-level image quality in seconds.
• This AI-driven method significantly reduces SPECT/CT reconstruction time, enabling faster clinical application.