A Reconstruction Method Based on Deep Convolutional Neural Network for SPECT Imaging

Chrysostomou, Charalambos; Koutsantonis, Loizos; Lemesios, Christos; Papanicolas, Costas N.

doi:10.1109/nssmic.2018.8824614

Cited by 11 publications

(3 citation statements)

References 18 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Although several previous studies applied large training datasets for deep learning-based improvement in SPECT imaging, these datasets were not realistic enough to be transferred to clinical SPECT systems. For example, Shao et al [23,24] and Chrysostomou et al [25] each used a large dataset consisting of analytically derived digital phantoms that showed limited physical effects such as photon scatter, attenuation, and non-perfect collimation. (Shao et al includes attenuation and non-perfect resolution in 2D.)…”

Section: Discussionmentioning

confidence: 99%

Analysis of a deep learning-based method for generation of SPECT projections based on a large Monte Carlo simulated dataset

et al. 2022

View full text Add to dashboard Cite

Background In recent years, a lot of effort has been put in the enhancement of medical imaging using artificial intelligence. However, limited patient data in combination with the unavailability of a ground truth often pose a challenge to a systematic validation of such methodologies. The goal of this work was to investigate a recently proposed method for an artificial intelligence-based generation of synthetic SPECT projections, for acceleration of the image acquisition process based on a large dataset of realistic SPECT simulations. Methods A database of 10,000 SPECT projection datasets of heterogeneous activity distributions of randomly placed random shapes was simulated for a clinical SPECT/CT system using the SIMIND Monte Carlo program. Synthetic projections at fixed angular increments from a set of input projections at evenly distributed angles were generated by different u-shaped convolutional neural networks (u-nets). These u-nets differed in noise realization used for the training data, number of input projections, projection angle increment, and number of training/validation datasets. Synthetic projections were generated for 500 test projection datasets for each u-net, and a quantitative analysis was performed using statistical hypothesis tests based on structural similarity index measure and normalized root-mean-squared error. Additional simulations with varying detector orbits were performed on a subset of the dataset to study the effect of the detector orbit on the performance of the methodology. For verification of the results, the u-nets were applied to Jaszczak and NEMA physical phantom data obtained on a clinical SPECT/CT system. Results No statistically significant differences were observed between u-nets trained with different noise realizations. In contrast, a statistically significant deterioration was found for training with a small subset (400 datasets) of the 10,000 simulated projection datasets in comparison with using a large subset (9500 datasets) for training. A good agreement between synthetic (i.e., u-net generated) and simulated projections before adding noise demonstrates a denoising effect. Finally, the physical phantom measurements show that our findings also apply for projections measured on a clinical SPECT/CT system. Conclusion Our study shows the large potential of u-nets for accelerating SPECT/CT imaging. In addition, our analysis numerically reveals a denoising effect when generating synthetic projections with a u-net. Clinically interesting, the methodology has proven robust against camera orbit deviations in a clinically realistic range. Lastly, we found that a small number of training samples (e.g., ~ 400 datasets) may not be sufficient for reliable generalization of the u-net.

show abstract

Section: Discussionmentioning

confidence: 99%

Analysis of a deep learning-based method for generation of SPECT projections based on a large Monte Carlo simulated dataset

et al. 2022

View full text Add to dashboard Cite

show abstract

“…Deep learning object detection is proposed to prevent collisions between detector heads and objects in the course of detector motion by creating a complete end-to-end deep learning object detector. (Chrysostomou et al, 2019)For SPECT imaging, we've come up with a novel way to rebuild tomographic images. The new "CNN Reconstruction CNNR" approach is based on deep convolutional neural networks (CNNs), which are utilised in the new reconstruction method.…”

Section: Review Of Literaturementioning

confidence: 99%

Artificial Intelligence & Deep Learning: Pet and Spect Imaging and Classification on Brain MRI Images List

Goel¹,

Yadav

Zubair

et al. 2022

J. Phys.: Conf. Ser.

View full text Add to dashboard Cite

Cognitive abilities commonly linked with human brains include learning and problem solving. Another option is just a preprogrammed rule that instructs the computer to perform a given action under specified circumstances. Voice recognition and image recognition, for instance, are both widely used in daily life and may be found everywhere. “Deep learning” (DL) approaches, in particular,due to their ability to come up with creative solutions to complex situations throughout the past decade. Both cancerous and non-cancerous tumors can be seen in the human brain (benign). Google developed the open-source deep learning library Tensor Flow. It can be combined with traditional machine learning. It was created on top of PIL that the Python Pillow module exists (Python Image Library). This collection of modules is critical to Python’s ability to perform image processing. “A deep learning API developed in Python” is powered by “the TensorFlow-based machine learning platform TensorFlow.” NumPy is a Python library for manipulating arrays, as the name suggests. Also included are linear algebra, the Fourier transform, and matrix functions. As far as the ResNet50 model goes, it has 48 convolutional layers, a maxpool and an averagepool layer. Up to a scale of 3.8 x 109, floating point operations can be performed. ResNet-50 is a type of neural network with 50 layers. You can import a pre-trained version of the network from the ImageNet database.

show abstract

“…Many alternative techniques apply Artificial Intelligence (AI), particularly Deep Learning (DL), to address the challenging task of tomographic image reconstruction from a set of noisy projections [5,[12][13][14][15][16][17][18]. These techniques utilize a deep neural network built and trained to provide the direct mapping between the projection space and image space.…”

Section: Introductionmentioning

confidence: 99%

A Data-Driven Reconstruction Technique based on Newton's Method for Emission Tomography

Koutsantonis¹,

Carneiro²,

Kieffer³

et al. 2021

Preprint

Self Cite

View full text Add to dashboard Cite

In this work, we present the Deep Newton Reconstruction Network (DNR-Net), a hybrid data-driven reconstruction technique for emission tomography inspired by Newton's method, a well-known iterative optimization algorithm. The DNR-Net employs prior information about the tomographic problem provided by the projection operator while utilizing deep learning approaches to a) imitate Newton's method by approximating the Newton descent direction and b) provide data-driven regularisation. We demonstrate that DNR-Net is capable of providing high-quality image reconstructions using data from SPECT phantom simulations by applying it to reconstruct images from noisy sinograms, each one containing 24 projections. The Structural Similarity Index (SSIM) and the Contrast-to-Noise ratio (CNR) were used to quantify the image quality. We also compare our results to those obtained by the OSEM method. According to the quantitative results, the DNR-Net produces reconstructions comparable to the ones produced by OSEM while featuring higher contrast and less noise.

show abstract

A Reconstruction Method Based on Deep Convolutional Neural Network for SPECT Imaging

Cited by 11 publications

References 18 publications

Analysis of a deep learning-based method for generation of SPECT projections based on a large Monte Carlo simulated dataset

Analysis of a deep learning-based method for generation of SPECT projections based on a large Monte Carlo simulated dataset

Artificial Intelligence & Deep Learning: Pet and Spect Imaging and Classification on Brain MRI Images List

A Data-Driven Reconstruction Technique based on Newton's Method for Emission Tomography

Contact Info

Product

Resources

About