Approximating anatomically-guided PET reconstruction in image space using a convolutional neural network

Schramm, Georg; Rigie, David; Vahle, Thomas; Rezaei, Ahmadreza; Laere, Koen Van; Shepherd, Timothy M.; Nuyts, Johan; Boada, Fernando E.

doi:10.1016/j.neuroimage.2020.117399

Cited by 40 publications

(30 citation statements)

References 37 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…If an image has one or more contours associated with it, the same transformation is applied to the contours. Geometric transformations are so common that they were utilised by 92 of the 93 basic augmentation studies 15–106 …”

Section: Methodsmentioning

confidence: 99%

See 1 more Smart Citation

A review of medical image data augmentation techniques for deep learning applications

et al. 2021

View full text Add to dashboard Cite

Summary Research in artificial intelligence for radiology and radiotherapy has recently become increasingly reliant on the use of deep learning‐based algorithms. While the performance of the models which these algorithms produce can significantly outperform more traditional machine learning methods, they do rely on larger datasets being available for training. To address this issue, data augmentation has become a popular method for increasing the size of a training dataset, particularly in fields where large datasets aren’t typically available, which is often the case when working with medical images. Data augmentation aims to generate additional data which is used to train the model and has been shown to improve performance when validated on a separate unseen dataset. This approach has become commonplace so to help understand the types of data augmentation techniques used in state‐of‐the‐art deep learning models, we conducted a systematic review of the literature where data augmentation was utilised on medical images (limited to CT and MRI) to train a deep learning model. Articles were categorised into basic, deformable, deep learning or other data augmentation techniques. As artificial intelligence models trained using augmented data make their way into the clinic, this review aims to give an insight to these techniques and confidence in the validity of the models produced.

show abstract

Section: Methodsmentioning

confidence: 99%

“…Gamma correction (3j), 107 linear contrast (3k) and histogram equalisation (3l) are common methods to adjust the contrast of an image. Twenty‐eight studies utilised intensity operations for data augmentation 18,20,21,28,34,38,40–42,45,46,55,62,67,70,75–77,79,84,87,88,94,102–104,106,108 …”

Section: Methodsmentioning

confidence: 99%

A review of medical image data augmentation techniques for deep learning applications

et al. 2021

View full text Add to dashboard Cite

show abstract

“…No pretraining pairs were needed; in fact, only the patient's prior information (T1-weighted MR) is needed, which is an unsupervised framework. Schramm et al (111) used 3D OSEM PET and 3D structure MRI as input to train a residual network (a purely convolutional shift-invariant neural network). Interestingly, their network has achieved good performance on tracer data that has never been seen before, proving that the network has better learned the denoising operation of the input PET image.…”

Section: Anatomical Image-guided Nuclear Medicine Image Reconstructionmentioning

confidence: 99%

Applications of artificial intelligence in nuclear medicine image generation

Cheng¹,

Wen²,

Huang³

et al. 2021

Quant Imaging Med Surg

View full text Add to dashboard Cite

Recently, the application of artificial intelligence (AI) in medical imaging (including nuclear medicine imaging) has rapidly developed. Most AI applications in nuclear medicine imaging have focused on the diagnosis, treatment monitoring, and correlation analyses with pathology or specific gene mutation.It can also be used for image generation to shorten the time of image acquisition, reduce the dose of injected tracer, and enhance image quality. This work provides an overview of the application of AI in image generation for single-photon emission computed tomography (SPECT) and positron emission tomography (PET) either without or with anatomical information [CT or magnetic resonance imaging (MRI)]. This review focused on four aspects, including imaging physics, image reconstruction, image postprocessing, and internal dosimetry. AI application in generating attenuation map, estimating scatter events, boosting image quality, and predicting internal dose map is summarized and discussed.

show abstract

“…Kim et al (27) generated HR SPECT images from LR SPECT images using a dense, block-based CNN. Furthermore, magnetic resonance-guided PET image reconstruction technology (mainly post-processing) can provide more prior details to improve the resolution of anatomical images (28); however, multi-modality images with good registration are not easy to obtain.…”

Section: Introductionmentioning

confidence: 99%

Super-resolution reconstruction for parallel-beam SPECT based on deep learning and transfer learning: a preliminary simulation study

Cheng¹,

Wen²,

Zhang³

et al. 2022

Ann Transl Med

View full text Add to dashboard Cite

Background: Single-photon emission computed tomography (SPECT) is widely used in the early diagnosis of major diseases such as cardiovascular disease and cancer. High-resolution (HR) imaging requires HR projection data, which typically comes with high costs. This study aimed to obtain HR SPECT images based on a deep learning algorithm using low-resolution (LR) detectors.Methods: A super-resolution (SR) reconstruction network based on deep learning and transfer learning for parallel-beam SPECT was proposed. LR SPECT sinograms were converted into HR sinograms.Training data were designed and generated, including digital phantoms (128×128 pixels), HR sinograms (128×128 pixels), and LR sinograms (128×64 pixels). A series of random phantoms was first used for pretraining, and then the extended cardiac-torso (XCAT) phantom was used to fine-tune the parameters. The effectiveness of the method was evaluated using an unknown cardiac phantom. To simulate a wide range of noise levels, the total count levels of the projection were normalized to 1e7 (100%), 1e6 (10%), and 1e5 (5%). Finally, the training sets for different count levels were generated. Transfer learning was employed to accelerate the training. Results:The proposed network was validated on the simulation data sets using different Poisson noise levels. The quantitative values of the peak signal-to-noise ratio (PSNR) and structural similarity index measure (SSIM) indicators of the reconstructed images were improved compared to those recorded using the benchmark methods. Using the proposed method, an image resolution comparable to that of images reconstructed from the HR projection could be achieved.Conclusions: Based on deep learning and transfer learning, an SR reconstruction network in the projection domain of the parallel-beam SPECT was developed. The simulation results under a wide range of noise levels evidenced the potential of the proposed network to improve SPECT resolution for LR detector scanners.

show abstract

Approximating anatomically-guided PET reconstruction in image space using a convolutional neural network

Cited by 40 publications

References 37 publications

A review of medical image data augmentation techniques for deep learning applications

A review of medical image data augmentation techniques for deep learning applications

Applications of artificial intelligence in nuclear medicine image generation

Super-resolution reconstruction for parallel-beam SPECT based on deep learning and transfer learning: a preliminary simulation study

Contact Info

Product

Resources

About