Initial Investigation of Low-Dose SPECT-MPI via Deep Learning

Abstract

Reduction of radiation exposure in SPECT-myocardial perfusion imaging (MPI) is critically important. However, lowering radiation dose significantly degrades image quality. Many deep learning structures have been implemented for denoising of low-dose CT data. However, to the best of our knowledge, there are no studies implementing deep learning structures for denoising of low-dose SPECT-MPI images. This paper reports a deep learning method to denoise SPECT-MPI images inspired by the recent work on low-dose CT. The proposed method is a 3D convolutional neural network based on stacked convolutional autoencoders, which is trained to map low-dose images to standard-dose images. Low-dose data was simulated for 1/8 and 1/16 of standard clinical dose. The performance of the proposed method was quantified by the average correlation (across patients in a test set) between predicted images and standard-dose images and was compared to that obtained from conventional denoising methods (i.e., spatial post-filtering). Preliminary results show that image quality improves significantly in the low-dose studies over conventional noise reduction methods. In particular, at 1/16 of clinical dose, the proposed method achieves similar image quality to that from 1/8 dose with conventional denoising. This suggests that further dose reduction could be achieved if the proposed method is used for post-processing of reconstructed images.