CoChat
Sign Up
Cited by 552Medical Image Segmentation TechniquesArtificial Intelligence in Healthcare and EducationAdvanced Neural Network Applications

NiftyNet: a deep-learning platform for medical imaging

Open access
Gibson, Eli, Li, Wenqi, Sudre, Carole, Fidon, Lucas, Shakir, Dzhoshkun I., Wang, Guotai, Eaton-Rosen, Zach, Gray, Robert +9 more·2018·Computer Methods and Programs in Biomedicine
DOI 10.1016/j.cmpb.2018.01.025PMC PMC5869052
Open this paper in CoChat — free account

Abstract

BACKGROUND AND OBJECTIVES: Medical image analysis and computer-assisted intervention problems are increasingly being addressed with deep-learning-based solutions. Established deep-learning platforms are flexible but do not provide specific functionality for medical image analysis and adapting them for this domain of application requires substantial implementation effort. Consequently, there has been substantial duplication of effort and incompatible infrastructure developed across many research groups. This work presents the open-source NiftyNet platform for deep learning in medical imaging. The ambition of NiftyNet is to accelerate and simplify the development of these solutions, and to provide a common mechanism for disseminating research outputs for the community to use, adapt and build upon. METHODS: The NiftyNet infrastructure provides a modular deep-learning pipeline for a range of medical imaging applications including segmentation, regression, image generation and representation learning applications. Components of the NiftyNet pipeline including data loading, data augmentation, network architectures, loss functions and evaluation metrics are tailored to, and take advantage of, the idiosyncracies of medical image analysis and computer-assisted intervention. NiftyNet is built on the TensorFlow framework and supports features such as TensorBoard visualization of 2D and 3D images and computational graphs by default. RESULTS: We present three illustrative medical image analysis applications built using NiftyNet infrastructure: (1) segmentation of multiple abdominal organs from computed tomography; (2) image regression to predict computed tomography attenuation maps from brain magnetic resonance images; and (3) generation of simulated ultrasound images for specified anatomical poses. CONCLUSIONS: The NiftyNet infrastructure enables researchers to rapidly develop and distribute deep learning solutions for segmentation, regression, image generation and representation learning applications, or extend the platform to new applications.

AI snapshot

Written from the paper’s full text.

TL;DR

NiftyNet is an open-source TensorFlow-based platform for deep learning in medical imaging, providing modular components for data loading, sampling, augmentation, networks, losses, evaluation, and a model zoo. The paper demonstrates it on abdominal CT segmentation, MRI-to-CT regression, and ultrasound simulation with a GAN.

Key findings

  • NiftyNet supports image segmentation, image regression, image model representation via auto-encoders, and image generation via auto-encoders and GANs, with modular components intended to reduce duplicated infrastructure effort.
  • In abdominal multi-organ CT segmentation using 90 CT scans and 9-fold cross-validation, median Dice scores ranged from 0.62 for duodenum to 0.95 for liver; other reported medians included spleen 0.94, left kidney 0.93, pancreas 0.75, and stomach 0.87.
  • In MRI-to-CT regression with 5-fold cross-validation, NiftyNet achieved mean absolute error of 88 HU (S.D. 7.5), compared with 121 HU (17) for pCT and 203 HU (24) for UTE; mean error was 9.1 HU (12) for NiftyNet, −7.3 HU (23) for pCT, and −132 HU (34) for UTE.
  • A pre-trained conditional GAN for ultrasound simulation was ported into the NiftyNet model zoo; interpolation results showed smooth variation in shadowing artifacts in one example and a sharp transition suggesting mode collapse in another.

Methods

  • NiftyNet is built on TensorFlow and comprises modular components including an ApplicationDriver, Application classes, Reader, Sampler, Network, and output handlers, with configuration via command line or human-readable configuration files and a model zoo for network reuse.
  • Three illustrative applications are presented: abdominal CT segmentation using DenseVNet, 90 CT scans, 9-fold cross-validation, 3000 iterations on whole images with ResizeSampler and random affine augmentation; MRI-to-CT regression using highresnet, 5-fold cross-validation, RMSE loss, 15,000 iterations on 80×80×80 patches with iSampler and random affine augmentation; and a conditional GAN for ultrasound simulation originally trained on 26,000 frames and ported via restore_initializer.
  • Segmentation and regression metrics were computed using NiftyNet’s evaluation action and aggregated over folds; segmentation metrics included Dice score, relative volume difference, mean absolute distance, and 95th percentile Hausdorff distance, while regression metrics included mean absolute error and mean error.

Limitations

  • Stated: NiftyNet currently supports a simplified machine learning pipeline that trains a single network and relies on users for data partitioning and model selection, such as hyper-parameter tuning; built-in cross-validation and standardized hyper-parameter tuning are future work.
  • Stated: the current model zoo contains a small number of models as proof of concept, and expanding it to state-of-the-art models for common tasks and public challenges is a future direction.
  • Evident: the GAN example was ported from prior work and evaluated qualitatively, with the authors noting a sharp transition suggesting mode collapse; the illustrative applications are specific examples rather than broad benchmarking across tasks or datasets.

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

Crossref · 10.1016/j.cmpb.2018.01.025

Cites

60 of 60 references verified0 citation errors · 0 not checkable

Statistics

Nothing to recompute.

Claims the paper makes

0 of 11 cited sources checked · 0 supported · 0 flagged

  • Findingreasoning

    NiftyNet's modular implementation facilitates adapting existing neural network architectures to new problems.

    “A modular implementation of the typical medical imaging machine learning pipeline facilitates (1) warm starts with established pre-trained networks, (2) adapting existing neural network architectures to new problems, and (3) rapid prototyping of new solutions.”
    Quote in the paperSubstantial claim
  • Findingreasoning

    NiftyNet's modular implementation of the typical medical imaging machine learning pipeline facilitates warm starts with established pre-trained networks.

    “A modular implementation of the typical medical imaging machine learning pipeline facilitates (1) warm starts with established pre-trained networks, (2) adapting existing neural network architectures to new problems, and (3) rapid prototyping of new solutions.”
    Quote in the paperSubstantial claim
  • Findingnone given

    NiftyNet is an open-source platform implemented based on TensorFlow APIs for deep learning in medical imaging.

    “An open-source platform is implemented based on TensorFlow APIs for deep learning in medical imaging domain.”
    Quote in the paperSubstantial claim
  • Findingreasoning

    NiftyNet's modular implementation facilitates rapid prototyping of new solutions.

    “A modular implementation of the typical medical imaging machine learning pipeline facilitates (1) warm starts with established pre-trained networks, (2) adapting existing neural network architectures to new problems, and (3) rapid prototyping of new solutions.”
    Quote in the paperSubstantial claim
  • Findingreasoning

    NiftyNet's conceptual blocks naturally encode a hierarchical structure for TensorBoard visualization.

    “because these conceptual blocks are constructed hierarchically through the composition of Layer objects and scopes, they naturally encode a hierarchical structure for TensorBoard visualization”
    Quote in the paperNot a substantial claim
  • Classified as probably not a substantial claim. 21
  • The quote only partly states this claim. 6
  • May combine two claims. 21

  • Probably not a substantial claim: may not pass S3 (load-bearing). The quote only partly states this claim.

    Findingreasoning

    The NiftyNet platform comprises several modular components, including ApplicationDriver, Application, Reader, Sampler, Network, and output handlers.

    “The NiftyNet platform comprises several modular components.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingnone given

    NiftyNet supports easy installation via pip (pip install niftynet).

    “The platform supports easy installation via the pip installation tool5 (i.e. pip install niftynet)”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingnone given

    The NiftyNet platform comprises an implementation of common infrastructure and common networks used in medical imaging, a database of pre-trained networks, and tools to facilitate adaptation of deep learning research to new clinical applications.

    “The NiftyNet platform comprises an implementation of the common infrastructure and common networks used in medical imaging, a database of pre-trained networks for specific applications and tools to facilitate the adaptation of deep learning research to new clinical applications with a shallow learning curve.”
    Quote in the paperSubstantial claim

  • May combine two claims. The quote only partly states this claim.

    Findingreasoning

    NiftyNet provides Network and Layer classes for constructing and reusing network architectures, with support for visualization via TensorBoard.

    “In NiftyNet, these conceptual blocks are represented by encapsulated Layer classes, or inline using TensorFlow’s scoping system.”
    • Also stated in: § Component details: networks and layers · plain
      “This also enables automatic support for visualization of the network graph as a hierarchy at different levels of detail using the TensorBoard visualizer [37] as shown in Fig. 3.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [37] D. Mané, et al., TensorBoard: TensorFlow’s visualization toolkit, 2015, https://github.com/tensorflow/tensorboard.

      Not checked in the free run.

  • May combine two claims. The quote only partly states this claim.

    Findingown data

    The NiftyNet Dense V-network achieved median Dice scores for 8 abdominal organs in 9-fold cross-validation, with the lowest median Dice score for duodenum (0.62) and the highest for liver (0.95).

    “Table 1 Median segmentation metrics for 8 organs aggregated over the 9-fold cross-validation.”
    • Also stated in: § Abdominal organ segmentation · plain
      “| Duodenum | 0.62 | 0.24 | 3.05 | 12.47 |”
    • Also stated in: § Abdominal organ segmentation · plain
      “| Liver | 0.95 | 0.02 | 1.42 | 4.12 |”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims. The quote only partly states this claim.

    Findingreasoning

    NiftyNet provides a Reader for loading medical image files and applying image-wide preprocessing, and a Sampler for generating data packets with augmentation.

    “The Reader class is responsible for loading corresponding image files from medical file formats for a specified data set, and applying image-wide preprocessing.”
    • Also stated in: § Component details: samplers and output handlers · plain
      “Sampler objects generate a sequence of packets of corresponding data for processing.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Findingnone given

    NiftyNet provides data normalization and augmentation layers, including mean, variance, histogram normalization, and flip, rotation, scaling augmentation.

    “NiftyNet currently supports mean, variance and histogram intensity data normalization, and flip, rotation and scaling spatial data augmentation.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingnone given

    NiftyNet provides analysis pipelines that can be run as part of the command line interface.

    “provides analysis pipelines that can be run as part of the command line interface”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingnone given

    NiftyNet releases encode the range of compatible versions of dependencies in a requirements.txt file, enabling automatic installation of compatible libraries for any NiftyNet version.

    “the range of compatible versions of NiftyNet dependencies will be encoded in a requirements.txt file in the code repository enabling automatic installation of compatible libraries for any NiftyNet version”
    Quote in the paperNot a substantial claim

  • May combine two claims. The quote only partly states this claim.

    Findingreasoning

    NiftyNet provides a model zoo and mechanisms for restoring, adapting, and reusing trained networks and conceptual blocks.

    “NiftyNet provides several mechanisms to support the distribution and reuse of networks and conceptual blocks.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingnone given

    NiftyNet currently supports image segmentation, image regression, image model representation, and image generation, and is designed to support addition of new application types.

    “NiftyNet currently supports •image segmentation,•image regression,•image model representation (via auto-encoder applications), and•image generation (via auto-encoder and generative adversarial networks (GANs)), and it is designed in a modular way to support the addition of new application types, by encapsulating typical application workflows in Application classes.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingcitation

    NiftyNet versions follow the semantic versioning 2.0 standard to ensure clear communication regarding backwards compatibility.

    “NiftyNet versions will follow the semantic versioning 2.0 standard [45] to ensure clear communication regarding backwards compatibility.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [45] T. Preston-Werner, Semantic versioning, Technical Report, T. Preston-Werner, 2015. URL: http://semver.org/.

      Not checked in the free run.

  • May combine two claims.

    Findingreasoning

    NiftyNet provides a high-level deep learning pipeline with components optimized for medical imaging applications and specific interfaces for medical image segmentation, classification, regression, image generation and representation learning applications.

    “NiftyNet provides a high-level deep learning pipeline with components optimized for medical imaging applications (data loading, sampling and augmentation, networks, loss functions, evaluations, and a model zoo) and specific interfaces for medical image segmentation, classification, regression, image generation and representation learning applications.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims. The quote only partly states this claim.

    Findingreasoning

    NiftyNet provides output handlers for training (loss and optimizer) and inference (aggregation and postprocessing), and supports evaluation metrics for segmentation.

    “During training, the output handler takes the network output, computes a loss and the gradient of the loss with respect to the trainable variables, and uses an Optimizer to iteratively train the model.”
    • Also stated in: § Component details: samplers and output handlers · plain
      “NiftyNet currently supports Aggregator objects for combining image patches, resizing images, and computing evaluation metrics.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingreasoning

    NiftyNet's conceptual blocks map to a meaningful subgraph of the computation graph and all associated variables, meaning that all weights for a conceptual block can be loaded into a new model with a single scope reference.

    “The scope of each conceptual blocks maps to a meaningful subgraph of the computation graph and all associated variables, meaning that all weights for a conceptual block can be loaded into a new model with a single scope reference.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Findingnone given

    The NiftyNet DenseVNet network was trained using images in their original NIfTI medical image format and the resulting trained model was publicly deployed in the NiftyNet model zoo.

    “The network was trained using images in their original NIfTI medical image format and the resulting trained model was publicly deployed in the NiftyNet model zoo.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingnone given

    The NiftyNet platform is available from http://niftynet.io/.

    “The NiftyNet platform is available from http://niftynet.io/.”
    Quote in the paperNot a substantial claim

  • Classified as probably not a substantial claim.

    Findingnone given

    This work presents the open-source NiftyNet platform for deep learning in medical imaging.

    “This work presents the open-source NiftyNet platform for deep learning in medical imaging.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Findingnone given

    The NiftyNet source code can be accessed from the Git repository or installed as a Python library using pip install niftynet.

    “The source code can be accessed from the Git repository6 or installed as a Python library using pip install niftynet.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingnone given

    NiftyNet is licensed under an open-source Apache 2.0 license.

    “NiftyNet is licensed under an open-source Apache 2.0 license7.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data

    The NiftyNet method for synthetic CT generation from MRI achieved a mean error of 9.1 HU, compared to -7.3 HU for pCT and -132 HU for UTE.

    “| ME | Average | 9.1 | −7.3 | −132 |”
    • Also stated in: § Image regression · plain
      “| | S.D. | 12 | 23 | 34 |”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Interpretationreasoning

    Different types of error may have very different clinical impacts, motivating specialized loss functions and evaluation metrics.

    “Furthermore, different types of error may have very different clinical impacts, motivating specialized loss functions and evaluation metrics (e.g. spatially weighted segmentation metrics).”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S4 (based). May combine two claims.

    Interpretationnone given

    Much of the infrastructure for deep learning pipelines is recreated by each researcher or research group, and much of it depends on the application domain being addressed.

    “Much of this infrastructure is recreated by each researcher or research group undertaking a deep learning project, and much of it depends on the application domain being addressed.”
    Quote in the paperNot a substantial claim
    • Rests on an assumption: There is substantial overlap in software pipelines implemented by many research groups.

  • May combine two claims.

    Interpretationreasoning

    General purpose deep learning platforms are unable to provide specific functionality for medical image analysis, and adapting them for this domain requires substantial implementation effort.

    “However, by avoiding assumptions about the application to remain general, the platforms are unable to provide specific functionality for medical image analysis and adapting them for this domain of application requires substantial implementation effort.”
    Quote in the paperSubstantial claim
    • Rests on an assumption: Deep learning technology has limited support in current software infrastructure for medical image analysis and computer-assisted intervention.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    Applications in computer-assisted intervention where analysis results are used in real time have additional constraints on analysis latency.

    “Applications in computer-assisted intervention where analysis results are used in real time (e.g. [21], [24]) have additional constraints on analysis latency.”
    Quote in the paperNot a substantial claim0 of 2 sources checked
    • cites [21] Garcia-Peraza-Herrera L.C. Li W. Fidon L. Gruijthuijsen C. Devreker A. Attilakos G. Deprest J. Poorten E.V. Stoyanov D. Vercauteren T. Ourselin S. ToolNet: holistically-nested real-time segmentation of robotic surgical tools 2017 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), Vancouver, BC 2017 5717 5722

      Not checked in the free run.

    • cites [24] Gibson E. Robu M.R. Thompson S. Edwards P.E. Schneider C. Gurusamy K. Davidson B. Hawkes D.J. Barratt D.C. Clarkson M.J. Deep residual networks for automatic segmentation of laparoscopic videos of the liver Proceedings of the SPIE, Medical Imaging vol. 10135 2017 101351M

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Study design and conductnone given

    The NiftyNet repository uses continuous integration incorporating system and unit tests for regression testing.

    “The NiftyNet repository uses continuous integration incorporating system and unit tests for regression testing.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Interpretationreasoning

    Independent reimplementation of custom infrastructure results in substantial duplication of effort, poses a barrier to dissemination of research tools, and inhibits fair comparisons between competing methods.

    “Independent reimplementation of all of this custom infrastructure results in substantial duplication of effort, poses a barrier to dissemination of research tools and inhibits fair comparisons between competing methods.”
    Quote in the paperNot a substantial claim
    • Rests on an assumption: There is substantial overlap in software pipelines implemented by many research groups.

  • May combine two claims.

    Interpretationreasoning

    The pre-NiftyNet implementation was not conducive to distributing the code or the trained network, and lacked visualizations for monitoring segmentation performance during training.

    “The pre-NiftyNet implementation was not conducive to distributing the code or the trained network, and lacked visualizations for monitoring segmentation performance during training.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S4 (based).

    Premisenone given

    Medical image analysis differs from other domains where deep learning is applied due to characteristics of the data itself and the applications in which they are used.

    “Medical image analysis differs from other domains where deep learning is applied due to characteristics of the data itself, and the applications in which they are used.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    Deep learning has been applied to a wide range of medical image analysis tasks and anatomical sites.

    “Recent reviews [34], [51] have highlighted that deep learning has been applied to a wide range of medical image analysis tasks (segmentation, classification, detection, registration, image reconstruction, enhancement, etc.) across a wide range of anatomical sites (brain, heart, lung, abdomen, breast, prostate, musculature, etc.).”
    Quote in the paperNot a substantial claim0 of 2 sources checked
    • cites [34] Litjens G. Kooi T. Bejnordi B.E. Setio A.A.A. Ciompi F. Ghafoorian M. van der Laak J.A.W.M. van Ginneken B. Sánchez C.I. A survey on deep learning in medical image analysis Medical Image Analysis 42 2017 60 88 28778026 source

      Not checked in the free run.

    • cites [51] Shen D. Wu G. Suk H.-I. Deep learning in medical image analysis Ann. Rev. Biomed. Eng. 19 2017 221 248 28301734 PMC5479722 source

      Not checked in the free run.

  • May combine two claims.

    Premisereasoning

    Medical image analysis must deal with large class imbalances, motivating special loss functions.

    “Because small image features can have large clinical importance, and because some pathology is very rare but life-threatening, medical image analysis must deal with large class imbalances, motivating special loss functions [18], [40], [53].”
    Quote in the paperNot a substantial claim0 of 3 sources checked
    • cites [18] L. Fidon, W. Li, L.C. Garcia-Peraza-Herrera, J. Ekanayake, N. Kitchen, S. Ourselin, T. Vercauteren, Generalised Wasserstein Dice score for imbalanced multi-class segmentation using holistic convolutional networks, (2017). arXiv: 1707.00478.

      Not checked in the free run.

    • cites [40] Milletari F. Navab N. Ahmadi S.-A. V-Net: Fully convolutional neural networks for volumetric medical image segmentation Proceedings of the Fourth International Conference on 3D Vision (3DV) 2016 565 571

      Not checked in the free run.

    • cites [53] Sudre C.H. Li W. Vercauteren T. Ourselin S. Cardoso M.J. Generalised Dice overlap as a deep learning loss function for highly unbalanced segmentations Proceedings of MICCAI Workshop on Deep Learning in Medical Image Analysis (DLMIA) 2017 PMC7610921 34104926

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

    Interpretationreasoning

    Using smaller data sets increases the importance of data augmentation, regularization, and cross-validation to prevent over-fitting.

    “Using smaller data sets increases the importance of data augmentation, regularization, and cross-validation to prevent over-fitting.”
    Quote in the paperNot a substantial claim
    • Rests on an assumption: Typical medical image data sets remain small.

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Interpretationreasoning

    The choice of TensorFlow for NiftyNet reflects engineering concerns (cross-platform support, multi-GPU support, built-in visualization, installation without compilation, semantic versioning) and pragmatic concerns (larger number of users, industry support).

    “The choice of the TensorFlow framework over the many deep learning frameworks described above reflects both engineering concerns – including cross-platform support, multi-GPU support, built-in visualization tools, installation without compilation, and semantic versioning – as well as pragmatic concerns, such as its larger number of users and support from industry.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Interpretationreasoning

    NiftyNet infrastructure enables researchers to rapidly develop deep learning solutions for segmentation, regression, image generation and representation learning applications, or extend the platform to new applications.

    “This infrastructure enables researchers to rapidly develop deep learning solutions for segmentation, regression, image generation and representation learning applications, or extend the platform to new applications.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Interpretationreasoning

    NiftyNet's named conceptual blocks simplified the adaptation of weights from pre-trained models and the TensorBoard-based hierarchical visualization of the computation graphs.

    “NiftyNet’s named conceptual blocks also simplified the adaptation of weights from pre-trained models and the TensorBoard-based hierarchical visualization of the computation graphs.”
    Quote in the paperNot a substantial claim
35 of 35

  • The paper relies on this without support.

    Deep learning usage in medical image analysis has increased by orders of magnitude in the last five years.

    “their usage has increased by orders of magnitude in the last five years.”

  • The paper relies on this without support.

    There is substantial overlap in software pipelines implemented by many research groups.

    “there is substantial overlap in software pipelines implemented by many research groups.”

  • The paper relies on this without support.

    Deep learning technology has limited support in current software infrastructure for medical image analysis and computer-assisted intervention.

    “deep learning technology has limited support in current software infrastructure for medical image analysis and computer-assisted intervention.”

  • The paper relies on this without support.

    Data dimensionality encountered in medical image analysis and computer-assisted intervention typically ranges from 2D to 5D.

    “Data dimensionality encountered in medical image analysis and computer-assisted intervention typically ranges from 2D to 5D.”

  • The paper relies on this without support.

    Typical medical image data sets remain small.

    “typical data sets remain small.”
5 of 5

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

Each contradiction between two passages listed here was confirmed by an independent check.

  • Classification is described as both supported and under development

    Major

    The platform description says NiftyNet provides a specific interface for classification applications, whereas the future-direction section places image classification among applications still under development.

  • Organ list omits gallbladder and repeats stomach

    Minor

    The text names seven distinct organs, listing stomach twice and omitting gallbladder. Table 1 instead reports results for eight distinct organs, including gallbladder.

2 of 2

Cites

Flagged first, then by citations

5 of the 65 references couldn’t be matched to a known paper.

Cited by

Sorted by citations

Related

Authors
Gibson, E., Li, W., Sudre, C., Fidon, L., Shakir, D. I., Wang, G., Eaton-Rosen, Z., Gray, R., Doel, T., Hu, Y., Whyntie, T., Nachev, P., Modat, M., Barratt, D. C., Ourselin, S., Cardoso, M. J., & Vercauteren, T.
Affiliations
  • Wellcome / EPSRC Centre for Interventional and Surgical Sciences (GB)
  • University College London (GB)
  • National Hospital for Neurology and Neurosurgery (GB)
Corresponding author
Li, Wenqi
ORCID
Venue
Computer Methods and Programs in Biomedicine
Publisher
Elsevier BV
Published
2018-05
Volume
158
Pages
113-122
ISSN
0169-2607 (ISSN-L)
DOI
10.1016/j.cmpb.2018.01.025
PMC
PMC5869052
OpenAlex
W2755930428
Semantic Scholar
c5762e451bc86e4742e575543cd545117b08de70
Cited by
552
References
65
Citations by year
2026: 12 · 2025: 19 · 2024: 41 · 2023: 50 · 2022: 75 · 2021: 104 · 2020: 113 · 2019: 94 · 2018: 41 · 2017: 2
FWCI
43.75
Type
journal article
Language
en
Open access
Licence
Locations
  • Computer Methods and Programs in Biomedicine (publishedVersion, cc-by)
  • Europe PMC (PubMed Central) (publishedVersion, cc-by)
  • PubMed Central (publishedVersion, cc-by)
  • PubMed (publishedVersion)
  • arXiv (Cornell University) (submittedVersion)
  • Lirias (submittedVersion, cc-by-nc-nd)
  • UCL Discovery (University College London) (submittedVersion)
  • Research Portal (King's College London) (submittedVersion, cc-by)
  • arXiv (Cornell University)
Funders
  • Nvidia
  • Wellcome Trust (HICF-T4-275, 203145Z, WT101957, 106882, 101957/Z/13/Z, 97914, /Z/15/Z, 203145/Z/16/Z, 106882/Z/15/Z)
  • Cancer Research UK (19985, A19985, C28070/A19985)
  • National Institute for Health and Care Research (HICF-T4-275, EP/M020533/1)
  • Royal Society (RG160569)
  • University College London (203145Z/16/Z, EP/L016478/1)
  • University College London Hospitals NHS Foundation Trust
  • Engineering and Physical Sciences Research Council (EP/L016478/1, EP/K503745/1, EP/M020533/1, M020533/1, EP/K503745, WT101957, M020533, 203145/Z/16/Z)
  • Science and Technology Facilities Council
Topics
  • Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Medical Image Segmentation Techniques
  • Health Sciences › Medicine › Health Informatics › Artificial Intelligence in Healthcare and Education
  • Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Advanced Neural Network Applications
Keywords
deep learning platform, medical image analysis, computer-based intervention, open source software, medical image segmentation, data augmentation, TensorFlow, 3D image visualization, abdominal organ segmentation, brain MRI
MeSH terms
Abdomen, Machine Learning, Brain, Computer Simulation, Diagnostic Imaging, Humans, Image Processing, Computer-Assisted, Magnetic Resonance Imaging, Ultrasonography, Databases, Factual, Neural Networks, Computer

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

Claims read by deepseek-v4.1-flash under definition version 3 (extract-v3) on 2026-10-09.

Classified by jev-1.13-20260917 (gold-guideline-v0.2) on 2026-10-09.

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review