SPECT reconstruction with a trained regularizer using CT-side information: Application to ¹⁷⁷Lu SPECT imaging

Area of Science:

• Nuclear Medicine
• Medical Imaging
• Radiochemistry

Background:

• Low-count single-photon emission computed tomography (SPECT) presents challenges for accurate imaging, particularly for radionuclides like ¹⁷⁷Lu with low photon yields.
• Conventional reconstruction methods often struggle with noise and quantification in low-count SPECT, impacting theranostic applications like dosimetry-based treatment planning.
• There is a need for advanced image reconstruction techniques to improve the quality and efficiency of low-count SPECT imaging.

Purpose of the Study:

• To develop and evaluate a trained regularizer for model-based SPECT image reconstruction that incorporates CT-derived segmentation information.
• To improve the accuracy of activity quantification and reduce noise in low-count SPECT imaging.
• To demonstrate the potential of the enhanced SPECT approach for pre-therapy theranostic imaging and reduced scan times.

Main Methods:

• A trained regularizer was developed for SPECT reconstruction, integrating segmentation masks derived from CT images using a pre-trained neural network (nnUNet).
• The proposed method was trained using patient studies with ¹⁷⁷Lu DOTATATE and validated with phantom and patient datasets under simulated low-count pre-therapy imaging conditions.
• Performance was compared against standard unregularized expectation-maximization (EM) algorithms and conventional CT-regularization methods.

Main Results:

• The trained regularizer significantly outperformed standard EM algorithms and conventional CT-regularization techniques in low-count SPECT reconstruction.
• Marked improvements were observed in activity quantification accuracy, noise reduction, and root mean square error (RMSE).
• The method demonstrated enhanced image quality and reliability under challenging low-count conditions.

Conclusions:

• The proposed trained regularizer effectively leverages CT-side information for improved low-count SPECT reconstruction.
• This approach offers a promising solution for enhancing theranostic imaging, enabling more accurate dosimetry and potentially shorter scan times.
• The enhanced low-count SPECT technique has broad implications for various nuclear medicine applications, including post-therapy imaging and whole-body SPECT.