DeepClas4Bio: Connecting bioimaging tools with deep learning frameworks for image classification

Inés, Adrián; Domínguez, César; Heras, Jónathan; Mata, Eloy; Pascual, Vico

doi:10.1016/j.compbiomed.2019.03.026

Cited by 13 publications

(6 citation statements)

References 13 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Application to other methods where vascular features appear on a visible tissue background, such as in X-ray histology, may require preprocessing or modifications to the cycle reconstruction loss functions. Accessibility of modern image analysis techniques to users continues to improve due to the development of several open-source bioimage platforms 44,[52][53][54][55][56][57][58][59][60] , which allow life scientists to intuitively apply advanced algorithms to their own data. Following testing across other modalities, future integration of VAN-GAN with such platforms would be valuable for widespread uptake.…”

Section: Discussionmentioning

confidence: 99%

Unsupervised segmentation of 3D microvascular photoacoustic images using deep generative learning

Sweeney

Hacker

Lefebvre

et al. 2023

Preprint

View full text Add to dashboard Cite

Innovations in imaging hardware have led to a revolution in our ability to visualise vascular networks in 3D at high resolution. The segmentation of microvascular networks from these 3D image volumes and interpretation of their meaning in the context of physiological and pathological processes unfortunately remains a time consuming and error-prone task. Deep learning has the potential to solve this problem, but current supervised analysis frameworks require human-annotated ground truth labels. To overcome these limitations, we present an unsupervised image-to-image translation deep learning model called the vessel segmentation generative adversarial network (VAN-GAN). VAN-GAN integrates synthetic blood vessel networks that closely resemble real-life anatomy into its training process and learns to replicate the underlying physics of an imaging system in order to learn how to segment vasculature from 3D biomedical images. To demonstrate the potential of VAN-GAN, the framework was applied to the challenge of segmenting vascular networks from images acquired using mesoscopic photoacoustic imaging (PAI). With a variety of in silico, in vitro and in vivo pathophysiological data, including patient-derived breast cancer xenograft models, we show that VAN-GAN facilitates accurate and unbiased segmentation of 3D vascular networks from PAI volumes. By leveraging synthetic data to reduce the reliance on manual labelling, VAN-GAN lower the barriers to entry for high-quality blood vessel segmentation to benefit imaging studies of vascular structure and function.

show abstract

Section: Discussionmentioning

confidence: 99%

Unsupervised segmentation of 3D microvascular photoacoustic images using deep generative learning

Sweeney

Hacker

Lefebvre

et al. 2023

Preprint

View full text Add to dashboard Cite

show abstract

“…A direct convolution that used video motion or extended short-term memory networks could have been an option. Alternatively, the study of DeepClass4Bio API by Inés et al (2019) could have bridged the gap between the ImageJ2 and PyTorch. The algorithm by Wu (2020) could have been added via the bridge to ImageJ2.…”

Section: Discussionmentioning

confidence: 99%

“…The Java programming language was used in many situations to create a bridge between two applications. Indeed, Inés et al (2019) highlighted a lack of a software bridge between bio-tools, ImageJ, and DeepLearning4j. According to Kainz et al (2015), commercial image retrieval software did not have the extensibility advantage of open-source software, and licensing issues for commercial software were one of the biggest reasons for lack of use.…”

Section: The Benefits Of Th E Java Programming Languagementioning

confidence: 99%

An Exploration of a Virtual Connection for Researchers and Educators by Exploring Strategies Enterprise Information Systems Specialists Need to Integrate Novel Neural Network Algorithms Into an Imaging Application – A Design Science Study

Quindazzi,

Sambasivam

2022

InSITE Conference

View full text Add to dashboard Cite

Aim/Purpose: The problem statement in the proposed study focuses on what strategies enterprise information systems specialists need to integrate novel algorithms into an imaging application that had not yet been identified. The aim is to demonstrate that a cross-convolutional neural network can be implemented within the home laboratory – an exploration of a virtual connection for re-search. An analysis of the works provides the basis for future extensibility of the software application for ImageJ2. Background: The study was guided by the research question: What strategies do enterprise information systems specialists need to integrate novel algorithms into an imaging application? This study demonstrates how to bring a lab-tested application online within a home laboratory to further build upon those findings. Methodology: A conceptual analysis was utilized for the artifact’s creation within the umbrella of design science to aid in the data interpretation segment. A conceptual framework was developed to determine relevant subject matter, such as useful software applications and technological enhancements to an image application. A research sample was not used in this study. Contribution: This research contributed to the body of knowledge by using a cross-convolutional neural network to explore novel algorithms as an enterprise information system specialist and set up the imaging application called Im-ageJ2 for development. The study’s setup is documented in a series of steps to demonstrate the how-to set up such a study. Findings: The findings were that it is possible to implement an existing work from within the home laboratory, steps of which are outlined for those to follow. Future work can be extended from the baseline workings. Furthermore, an analysis of the existing code was determined to see if the existing PyTorch code could be developed within Java to act as an extension to ImageJ2 later. By examining the programming code and the cross-convolutional functionality, a determination was made on the best Java mapping. A set of highlights of the processes used are included. Recommendations for Practitioners: The study intersects two differing realms: artificial intelligence and the enterprise information systems network. This study builds upon an AWS system and details the steps to implement an artificial intelligence system on the Amazon Web Service (AWS) platform. The researcher investigated existing soft-ware imaging Java applications and probed the areas of potential extensibility for implementing the artificial intelligence novel algorithms. Recommendations for Researchers: The concept of being able to bring online a turnkey set of computer hard-ware off-the-shelf within the home laboratory may be an unexplored avenue to some researchers. The recommendation is to encourage researchers to see past the constraints that a lack of hardware may bring about for computer science research. The researcher no longer has to be within an office laboratory on campus to access power computers. The doors are open to exploring a whole new world. It explores how to move through the various linkage is-sues with old libraries and new systems. Other avenues this research can enhance are extending the ImageJ libraries as a Java plugin with the novel algorithms such as the cross-convolutional network implemented on the AWS platform. Impact on Society: Expanding the accessibility to the researcher and practitioner field could be profound. No longer is the researcher or practitioner constrained to the office laboratory. This work shows how to move through the issues to invoke the software to produce and investigate existing software. The findings and guidance of this study are profound. Future Research: Future research should focus on implementing the mappings table as deter-mined in previous research. A future design could be pursued from the analysis and implementation of a Java neural network algorithm as a basic implementation. Java can chain elements to begin replicating the algorithm by Xue et al. (2019). The cross-convolutional element should have further analysis to ensure the full replication within Java. A cross-convolutional neural network (CCNN) Java algorithm could be implemented and trial run within the code; this would allow for the leverage of the heavy lifting of the program code to predict image motion. There seemed to be few approaches regarding predicting motion frames of a future image, and such predictions could be a giant leap for the health care world regarding microscopic and x-ray images.

show abstract

“…Different computational tools produce different output because they rely on different approaches, from conventional mathematical operations designed to filter and segment objects, to state-of-the-art deep learning neural networks [ 64 ] that identify objects based on hundreds of features or rules. Some deep-learning implementations are focused on tasks such as object detection [ 65 , 66 ], image segmentation [ 67 , 68 ], object tracking [ 69 , 70 ], object classification [ 71 ] or a combination of these [ 72 , 73 , 74 , 75 , 76 ].…”

Section: The Developing Vertebrate Retina In Three-dimensionsmentioning

confidence: 99%

Quantitative Approaches to Study Retinal Neurogenesis

Ledesma-Terrón

Míguez

2021

Biomedicines

View full text Add to dashboard Cite

The study of the development of the vertebrate retina can be addressed from several perspectives: from a purely qualitative to a more quantitative approach that takes into account its spatio-temporal features, its three-dimensional structure and also the regulation and properties at the systems level. Here, we review the ongoing transition toward a full four-dimensional characterization of the developing vertebrate retina, focusing on the challenges at the experimental, image acquisition, image processing and quantification. Using the developing zebrafish retina, we illustrate how quantitative data extracted from these type of highly dense, three-dimensional tissues depend strongly on the image quality, image processing and algorithms used to segment and quantify. Therefore, we propose that the scientific community that focuses on developmental systems could strongly benefit from a more detailed disclosure of the tools and pipelines used to process and analyze images from biological samples.

show abstract

DeepClas4Bio: Connecting bioimaging tools with deep learning frameworks for image classification

Cited by 13 publications

References 13 publications

Unsupervised segmentation of 3D microvascular photoacoustic images using deep generative learning

Unsupervised segmentation of 3D microvascular photoacoustic images using deep generative learning

An Exploration of a Virtual Connection for Researchers and Educators by Exploring Strategies Enterprise Information Systems Specialists Need to Integrate Novel Neural Network Algorithms Into an Imaging Application – A Design Science Study

Quantitative Approaches to Study Retinal Neurogenesis

Contact Info

Product

Resources

About