Improvements of 111in Spect Images Reconstructed With Sparsely Acquired Projections by Deep Learning Generated Synthetic Projections

Abstract

The aim was to improve single-photon emission computed tomography (SPECT) quality for sparsely acquired 111In projections by adding deep learning generated synthetic intermediate projections (SIPs). Method: The recently constructed deep convolutional network for generating synthetic intermediate projections (CUSIP) was used for improving 20 sparsely acquired 111In-octreotide SPECTs. Reconstruction was performed with 120 (120P) or 30 (30P) projections, or 120 projections with 90 SIPs generated from 30 projections (30–120SIP). The SPECT reconstructions were performed with attenuation, scatter and collimator response corrections. Postfiltered 30P reconstructed SPECT was also analyzed. Image quality were quantitatively evaluated with root-mean-square error, peak signal-to-noise ratio and structural similarity index metrics. Result: The 30–120SIP reconstructed SPECT had statistically significant improved image quality parameters compared to 30P reconstructed SPECT with and without post filtering. The images visual appearance was similar to slightly filtered 120P SPECTs. Thereby, substantial acquisition time reduction with SIPs seems possible without image quality degradation.

Improvements of 177Lu SPECT images from sparsely acquired projections by reconstruction with deep-learning-generated synthetic projections

“…The use of SIPs seems to be a promising noise reduction technique, as it has previously yielded improved results compared to regular post filtering with Butterworth or Gaussian filters [ 16 , 17 ]. Future studies should be conducted with the aim of examining the detectability of small lesions in SPECT/CT imaging using SIPs instead of standard filtering.…”

“…The deep neural network was designed and trained to generate 90 SIPs from an input of 30P. This network has previously been shown to perform well in reconstructions of 177 Lu-DOTATATE and Indium-111 ( 111 In)-octreotide images, showing high structural similarity between acquired projections and SIPs, and improved image quality in reconstructed images compared to reconstruction of 30P [ 16 , 17 ]. However, this study only evaluated images from one day post administration and did not evaluate any dosimetry.…”

Evaluation of reconstruction methods and image noise levels concerning visual assessment of simulated liver lesions in 111In-octreotide SPECT imaging

“…It has previously been shown that MC OSEM signi cantly improves the image quality in 177 Lu-octreotate imaging [21] and the spatial resolution in 111 In-octreotide imaging [15]. However, the noise level needs to be handled appropriately, and we aim to further investigate deep learning-generated synthetic intermediate projections (SIPs) in SPECT images, which have been demonstrated to more effectively reduce the noise level compared to post-ltering methods such as Gaussian ltering [23,24]. This might improve SNR in images reconstructed with MC-based OSEM reconstruction.…”

Evaluation of reconstruction methods and image noise levels concerning visual assessment of simulated liver lesions in 111In-octreotide SPECT imaging

“…Hence, MC-based reconstruction is very promising. However, the noise level needs to be handled appropriately, and we aim to further investigate deep learning–generated synthetic intermediate projections (SIPs) in SPECT images, which have been demonstrated to more effectively reduce the noise level compared to post-filtering methods such as Gaussian filtering [ 32 , 33 ]. This might improve SNR in images reconstructed with MC-based OSEM reconstruction.…”