Comparison and Analysis of Image-to-Image Generative Adversarial Networks: A Survey

Saxena, Sagar; Teli, Mohammad Nayeem

doi:10.48550/arxiv.2112.12625

Cited by 3 publications

(3 citation statements)

References 25 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…89 In the DL literature, many supervised and unsupervised image translation architectures exist, but only a small subset have been applied to medical data. A more complete review of these architectures can be found in Alotaibi et al 90 and Pang et al, 91 with benchmarks of various networks on some of the most common datasets in Saxena et al 92 Interestingly, unpaired training configurations outperformed paired training for cycle-GAN in terms of relative MAE improvements (see Figure 6; 55.98% vs. 47.61%, p = 0.16). However, the results narrowed when studies with similar training set sizes were compared,with paired implementations performing slightly better (53.65% vs. 51.83%, p = 0.29).…”

Section: Recommendations For Researchersmentioning

confidence: 95%

Deep learning methods for enhancing cone‐beam CT image quality toward adaptive radiation therapy: A systematic review

et al. 2022

View full text Add to dashboard Cite

The use of deep learning (DL) to improve cone‐beam CT (CBCT) image quality has gained popularity as computational resources and algorithmic sophistication have advanced in tandem. CBCT imaging has the potential to facilitate online adaptive radiation therapy (ART) by utilizing up‐to‐date patient anatomy to modify treatment parameters before irradiation. Poor CBCT image quality has been an impediment to realizing ART due to the increased scatter conditions inherent to cone‐beam acquisitions. Given the recent interest in DL applications in radiation oncology, and specifically DL for CBCT correction, we provide a systematic theoretical and literature review for future stakeholders. The review encompasses DL approaches for synthetic CT generation, as well as projection domain methods employed in the CBCT correction literature. We review trends pertaining to publications from January 2018 to April 2022 and condense their major findings—with emphasis on study design and DL techniques. Clinically relevant endpoints relating to image quality and dosimetric accuracy are summarized, highlighting gaps in the literature. Finally, we make recommendations for both clinicians and DL practitioners based on literature trends and the current DL state‐of‐the‐art methods utilized in radiation oncology.

show abstract

Section: Recommendations For Researchersmentioning

confidence: 95%

Deep learning methods for enhancing cone‐beam CT image quality toward adaptive radiation therapy: A systematic review

et al. 2022

View full text Add to dashboard Cite

show abstract

“…The generator takes an input image and generates an output image, while the discriminator differentiates between the real and fake output images. The translation processes from one domain to the other were translated numerically as the generator loss values, which are considered the objective equivalent of determining whether the model was implemented correctly (Saxena, Teli, 2021). To generate new architectural images from a sound file, the pix2pix algorithm was trained with spectrogram files as source files, and architectural images as target files.…”

Section: Pix2pix Based Modelmentioning

confidence: 99%

Utilizing Web Applications for Developing Digital Twin: Case study urban tunnel with kinetic roof

Rezakhani,

Kim

2023

eCAADe Proceedings

View full text Add to dashboard Cite

The development of novel design output using Artificial Neural Networks (ANNs) is becoming an important milestone in the architectural design discourse. With the recent encounter of the computational design realm with the diffusion models, it becomes even easier to generate 2D and 3D design outputs. Yet, the utilization of machine learning tools within design computing domains is confined to generating or classifying visual and encoded data. However, it is critical to evaluate the untapped potentials of machine learning technologies in terms of illuminating the implicit correlations and links underlying distinct concepts and themes across a wide range of technical domains. With the ongoing research project named "Local Intelligence", we hypothesized that the local knowledge of a certain location might be conceptualized as a distributed network to connect different forms of local knowledge. As the first case of the project, we tried to reinstate a commonality between the local music and vernacular architecture, for which we trained generative adversarial network (GAN) models with the visual spectrograms translated from the audio data of the local songs and images of vernacular architectural instances from a defined geography. The two multi-modal GAN models differ in terms of the inherent convolutional layers and data pairing process. The outcomes demonstrated that both GAN models can learn how to depict vernacular architectural features from the rhythmic pattern of the songs in various patterns. Consequently, the implicit relations between music and architecture in the initial findings come one step closer to being demystified. Thus, the process and generative outcomes of the two models are compared and discussed in terms of the legibility of the architectural features, by taking the original vernacular architectural image dataset as the ground truth.

show abstract

“…One type of Generative Adversarial Network, a CycleGAN, is designed for image translation [16] and can alleviate this issue. It is trained using an adversarial process involving generators and discriminators, but a CycleGAN has two pairs.…”

Section: Introductionmentioning

confidence: 99%

Ensuring accurate stain reproduction in deep generative networks for virtual immunohistochemistry

Walsh¹,

Edwards²,

Insall³

2022

Preprint

View full text Add to dashboard Cite

Immunohistochemistry is a valuable diagnostic tool for cancer pathology. However, it requires specialist labs and equipment, is time-intensive, and is difficult to reproduce. Consequently, a long term aim is to provide a digital method of recreating physical immunohistochemical stains. Generative Adversarial Networks have become exceedingly advanced at mapping one image type to another and have shown promise at inferring immunostains from haematoxylin and eosin. However, they have a substantial weakness when used with pathology images as they can fabricate structures that are not present in the original data. CycleGANs can mitigate invented tissue structures in pathology image mapping but have a related disposition to generate areas of inaccurate staining. In this paper, we describe a modification to the loss function of a CycleGAN to improve its mapping ability for pathology images by enforcing realistic stain replication while retaining tissue structure. Our approach improves upon others by considering structure and staining during model training. We evaluated our network using the Fréchet Inception distance, coupled with a new technique that we propose to appraise the accuracy of virtual immunohistochemistry. This assesses the overlap between each stain component in the inferred and ground truth images through colour deconvolution, thresholding and the Sorensen-Dice coefficient. Our modified loss function resulted in a Dice coefficient for the virtual stain of 0.78 compared with the real AE1/AE3 slide. This was superior to the unaltered CycleGAN's score of 0.74. Additionally, our loss function improved the Fréchet Inception distance for the reconstruction to 74.54 from 76.47. We, therefore, describe an advance in virtual restaining that can extend to other immunostains and tumour types and deliver reproducible, fast and readily accessible immunohistochemistry worldwide.

show abstract

Comparison and Analysis of Image-to-Image Generative Adversarial Networks: A Survey

Cited by 3 publications

References 25 publications

Deep learning methods for enhancing cone‐beam CT image quality toward adaptive radiation therapy: A systematic review

Deep learning methods for enhancing cone‐beam CT image quality toward adaptive radiation therapy: A systematic review

Utilizing Web Applications for Developing Digital Twin: Case study urban tunnel with kinetic roof

Ensuring accurate stain reproduction in deep generative networks for virtual immunohistochemistry

Contact Info

Product

Resources

About