Student or educator? Sign up with your school email and get 6 months of Pro free.
CoChat

The DOI matches its Crossref record; 65 of the 65 references it cites were found in a registry or an index; 30 of the 49 sources its claims cite were checked for support.

Convolutional neural networks: an overview and application in radiology

Open access
Yamashita, Rikiya, Nishio, Mizuho, Do, Richard Kinh Gian, Togashi, Kaori · 2018 · Insights into Imaging

Abstract

Convolutional neural network (CNN), a class of artificial neural networks that has become dominant in various computer vision tasks, is attracting interest across a variety of domains, including radiology. CNN is designed to automatically and adaptively learn spatial hierarchies of features through backpropagation by using multiple building blocks, such as convolution layers, pooling layers, and fully connected layers. This review article offers a perspective on the basic concepts of CNN and its application to various radiological tasks, and discusses its challenges and future directions in the field of radiology. Two challenges in applying CNN to radiological tasks, small dataset and overfitting, will also be covered in this article, as well as techniques to minimize them. Being familiar with the concepts and advantages, as well as limitations, of CNN is essential to leverage its potential in diagnostic radiology, with the goal of augmenting the performance of radiologists and improving patient care. KEY POINTS: • Convolutional neural network is a class of deep learning methods which has become dominant in various computer vision tasks and is attracting interest across a variety of domains, including radiology. • Convolutional neural network is composed of multiple building blocks, such as convolution layers, pooling layers, and fully connected layers, and is designed to automatically and adaptively learn spatial hierarchies of features through a backpropagation algorithm. • Familiarity with the concepts and advantages, as well as limitations, of convolutional neural network is essential to leverage its potential to improve radiologist performance and, eventually, patient care.

Overview

CoChat Summary

TL;DR

This narrative review explains the basic concepts of convolutional neural networks (CNNs) — their building blocks, training, and data requirements — and surveys their application to radiological tasks such as classification, segmentation, detection, image denoising, and report analysis. It highlights small datasets and overfitting as the two central challenges for CNN use in radiology and reviews techniques to mitigate them.

Key findings

  • CNN is composed of convolution, pooling, and fully connected layers; convolution and pooling perform feature extraction while fully connected layers map features to outputs, with kernels and weights learned via backpropagation and gradient descent.
  • CNN differs from conventional radiomics in that it does not require hand-crafted feature extraction or expert segmentation, but is far more data hungry and computationally expensive, typically requiring GPUs.
  • Reported application results cited in the review include: an ensemble of AlexNet and GoogLeNet 2D-CNNs achieving a best AUC of 0.99 for detecting pulmonary tuberculosis on 1007 chest radiographs; a CNN-based CADe system trained on ~45,000 mammography images outperforming a hand-crafted-feature CADe system at low sensitivity and matching it at high sensitivity; and a detection system built from 33,688 bounding boxes mined from radiologists' reports showing significantly improved detection accuracy with pseudo-category labels.
  • Transfer learning from ImageNet (1.4 million images, 1000 classes) via fixed feature extraction or fine-tuning is presented as a key strategy for training on small datasets; U-net is highlighted for segmentation because it captures multiscale spatial context and can be trained end-to-end from limited data.

Methods

  • Design: narrative review article (not a primary study), covering CNN concepts, training, data/label considerations, overfitting, small-dataset strategies, and radiology applications.
  • Scope: applications are organized into classification, segmentation, detection, and "others" (denoising, image reconstruction, report classification); recurrent neural networks and other deep learning models are explicitly stated to be beyond the article's scope.
  • Evidence base: illustrative prior studies are cited rather than systematically searched or quantitatively pooled; no new data, sample, or statistical analysis is generated by the authors.

Limitations

  • Stated: the article is a perspective/overview, and other deep learning models such as recurrent neural networks for sequence models are beyond its scope; the details of improved gradient descent algorithms (momentum, RMSprop, Adam) are also stated to be beyond scope.
  • Stated: whether semi-supervised techniques (pseudo-label, GANs) actually improve deep learning performance in radiology "is not clear and remains an area of active investigation," and whether medical pretrained networks would outperform ImageNet-pretrained transfer learning is likewise unclear.
  • Evident from the text: as a narrative review it reports no systematic search strategy, inclusion criteria, or quantitative synthesis, so the cited performance figures are illustrative examples rather than a comparative or exhaustive assessment; the authors also note deep learning's "black box" nature and vulnerability to adversarial examples as unresolved concerns for clinical robustness.

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

Integrity signals

Registry

Crossref · 10.1007/s13244-018-0639-9

Cites

65 of 65 references verified0 citation errors · 0 not checkable

Statistics

Nothing to recompute.

Figures

Fig. 1 An overview of a convolutional neural network (CNN) architecture and the training process. A CNN is composed of a stacking of several building blocks: convolution layers, pooling layers (e.g., max pooling), and fully connected (FC) layers. A model’s performance under particular kernels and weights is calculated with a loss function through forward propagation on a training dataset, and learnable parameters, i.e., kernels and weights, are updated according to the loss value through backpropagation with gradient descent optimization algorithm. ReLU, rectified linear unit
Fig. 1 An overview of a convolutional neural network (CNN) architecture and the training process. A CNN is composed of a stacking of several building blocks: convolution layers, pooling layers (e.g., max pooling), and fully connected (FC) layers. A model’s performance under particular kernels and weights is calculated with a loss function through forward propagation on a training dataset, and learnable parameters, i.e., kernels and weights, are updated according to the loss value through backpropagation with gradient descent optimization algorithm. ReLU, rectified linear unitCC BY · PubMed Central
Fig. 2 A computer sees an image as an array of numbers. The matrix on the right contains numbers between 0 and 255, each of which corresponds to the pixel brightness in the left image. Both are overlaid in the middle image. The source image was downloaded via http://yann.lecun.com/exdb/mnist
Fig. 2 A computer sees an image as an array of numbers. The matrix on the right contains numbers between 0 and 255, each of which corresponds to the pixel brightness in the left image. Both are overlaid in the middle image. The source image was downloaded via http://yann.lecun.com/exdb/mnistCC BY · PubMed Central
Fig. 3 a–c An example of convolution operation with a kernel size of 3 × 3, no padding, and a stride of 1. A kernel is applied across the input tensor, and an element-wise product between each element of the kernel and the input tensor is calculated at each location and summed to obtain the output value in the corresponding position of the output tensor, called a feature map. d Examples of how kernels in convolution layers extract features from an input tensor are shown. Multiple kernels work as different feature extractors, such as a horizontal edge detector (top), a vertical edge detector (middle), and an outline detector (bottom). Note that the left image is an input, those in the middle are kernels, and those in the right are output feature maps
Fig. 3 a–c An example of convolution operation with a kernel size of 3 × 3, no padding, and a stride of 1. A kernel is applied across the input tensor, and an element-wise product between each element of the kernel and the input tensor is calculated at each location and summed to obtain the output value in the corresponding position of the output tensor, called a feature map. d Examples of how kernels in convolution layers extract features from an input tensor are shown. Multiple kernels work as different feature extractors, such as a horizontal edge detector (top), a vertical edge detector (middle), and an outline detector (bottom). Note that the left image is an input, those in the middle are kernels, and those in the right are output feature mapsCC BY · PubMed Central
Fig. 4 A convolution operation with zero padding so as to retain in-plane dimensions. Note that an input dimension of 5 × 5 is kept in the output feature map. In this example, a kernel size and a stride are set as 3 × 3 and 1, respectively
Fig. 4 A convolution operation with zero padding so as to retain in-plane dimensions. Note that an input dimension of 5 × 5 is kept in the output feature map. In this example, a kernel size and a stride are set as 3 × 3 and 1, respectivelyCC BY · PubMed Central
Fig. 5 Activation functions commonly applied to neural networks: a rectified linear unit (ReLU), b sigmoid, and c hyperbolic tangent (tanh)
Fig. 5 Activation functions commonly applied to neural networks: a rectified linear unit (ReLU), b sigmoid, and c hyperbolic tangent (tanh)CC BY · PubMed Central
Fig. 6 a An example of max pooling operation with a filter size of 2 × 2, no padding, and a stride of 2, which extracts 2 × 2 patches from the input tensors, outputs the maximum value in each patch, and discards all the other values, resulting in downsampling the in-plane dimension of an input tensor by a factor of 2. b Examples of the max pooling operation on the same images in Fig. 3b. Note that images in the upper row are downsampled by a factor of 2, from 26 × 26 to 13 × 13
Fig. 6 a An example of max pooling operation with a filter size of 2 × 2, no padding, and a stride of 2, which extracts 2 × 2 patches from the input tensors, outputs the maximum value in each patch, and discards all the other values, resulting in downsampling the in-plane dimension of an input tensor by a factor of 2. b Examples of the max pooling operation on the same images in Fig. 3b. Note that images in the upper row are downsampled by a factor of 2, from 26 × 26 to 13 × 13CC BY · PubMed Central
Fig. 7 Gradient descent is an optimization algorithm that iteratively updates the learnable parameters so as to minimize the loss, which measures the distance between an output prediction and a ground truth label. The gradient of the loss function provides the direction in which the function has the steepest rate of increase, and all parameters are updated in the negative direction of the gradient with a step size determined based on a learning rate
Fig. 7 Gradient descent is an optimization algorithm that iteratively updates the learnable parameters so as to minimize the loss, which measures the distance between an output prediction and a ground truth label. The gradient of the loss function provides the direction in which the function has the steepest rate of increase, and all parameters are updated in the negative direction of the gradient with a step size determined based on a learning rateCC BY · PubMed Central
Fig. 8 Available data are typically split into three sets: a training, a validation, and a test set. A training set is used to train a network, where loss values are calculated via forward propagation and learnable parameters are updated via backpropagation. A validation set is used to monitor the model performance during the training process, fine-tune hyperparameters, and perform model selection. A test set is ideally used only once at the very end of the project in order to evaluate the performance of the final model that is fine-tuned and selected on the training process with training and validation sets
Fig. 8 Available data are typically split into three sets: a training, a validation, and a test set. A training set is used to train a network, where loss values are calculated via forward propagation and learnable parameters are updated via backpropagation. A validation set is used to monitor the model performance during the training process, fine-tune hyperparameters, and perform model selection. A test set is ideally used only once at the very end of the project in order to evaluate the performance of the final model that is fine-tuned and selected on the training process with training and validation setsCC BY · PubMed Central
Fig. 9 A routine check for recognizing overfitting is to monitor the loss on the training and validation sets during the training iteration. If the model performs well on the training set compared to the validation set, then the model has been overfit to the training data. If the model performs poorly on both training and validation sets, then the model has been underfit to the data. Although the longer a network is trained, the better it performs on the training set, at some point, the network fits too well to the training data and loses its capability to generalize
Fig. 9 A routine check for recognizing overfitting is to monitor the loss on the training and validation sets during the training iteration. If the model performs well on the training set compared to the validation set, then the model has been overfit to the training data. If the model performs poorly on both training and validation sets, then the model has been underfit to the data. Although the longer a network is trained, the better it performs on the training set, at some point, the network fits too well to the training data and loses its capability to generalizeCC BY · PubMed Central
Fig. 10 Transfer learning is a common and effective strategy to train a network on a small dataset, where a network is pretrained on an extremely large dataset, such as ImageNet, then reused and applied to the given task of interest. A fixed feature extraction method is a process to remove FC layers from a pretrained network and while maintaining the remaining network, which consists of a series of convolution and pooling layers, referred to as the convolutional base, as a fixed feature extractor. In this scenario, any machine learning classifier, such as random forests and support vector machines, as well as the usual FC layers, can be added on top of the fixed feature extractor, resulting in training limited to the added classifier on a given dataset of interest. A fine-tuning method, which is more often applied to radiology research, is to not only replace FC layers of the pretrained model with a new set of FC layers to retrain them on a given dataset, but to fine-tune all or part of the kernels in the pretrained convolutional base by means of backpropagation. FC, fully connected
Fig. 10 Transfer learning is a common and effective strategy to train a network on a small dataset, where a network is pretrained on an extremely large dataset, such as ImageNet, then reused and applied to the given task of interest. A fixed feature extraction method is a process to remove FC layers from a pretrained network and while maintaining the remaining network, which consists of a series of convolution and pooling layers, referred to as the convolutional base, as a fixed feature extractor. In this scenario, any machine learning classifier, such as random forests and support vector machines, as well as the usual FC layers, can be added on top of the fixed feature extractor, resulting in training limited to the added classifier on a given dataset of interest. A fine-tuning method, which is more often applied to radiology research, is to not only replace FC layers of the pretrained model with a new set of FC layers to retrain them on a given dataset, but to fine-tune all or part of the kernels in the pretrained convolutional base by means of backpropagation. FC, fully connectedCC BY · PubMed Central
Fig. 11 A schematic illustration of a classification system with CNN and representative examples of its training data. a Classification system with CNN in the deployment phase. b, c Training data used in training phase
Fig. 11 A schematic illustration of a classification system with CNN and representative examples of its training data. a Classification system with CNN in the deployment phase. b, c Training data used in training phaseCC BY · PubMed Central
Fig. 12 A schematic illustration of the system for segmenting a uterus with a malignant tumor and representative examples of its training data. a Segmentation system with CNN in deployment phase. b Training data used in the training phase. Note that original images and corresponding manual segmentations are arranged next to each other
Fig. 12 A schematic illustration of the system for segmenting a uterus with a malignant tumor and representative examples of its training data. a Segmentation system with CNN in deployment phase. b Training data used in the training phase. Note that original images and corresponding manual segmentations are arranged next to each otherCC BY · PubMed Central
Fig. 13 A schematic illustration of the system for denoising an ultra-low-dose CT (ULDCT) image of phantom and representative examples of its training data. a Denoising system with CNN in deployment phase. b Training data used in training phase. SDCT, standard-dose CT
Fig. 13 A schematic illustration of the system for denoising an ultra-low-dose CT (ULDCT) image of phantom and representative examples of its training data. a Denoising system with CNN in deployment phase. b Training data used in training phase. SDCT, standard-dose CTCC BY · PubMed Central
Fig. 14 An example of a class activation map (CAM) [58]. A CNN network trained on ImageNet classified the left image as a “bridge pier”. A heatmap for the category of “bridge pier”, generated by a method called Grad-CAM [59], is superimposed (right image), which indicates the discriminative image regions used by the CNN for the classification
Fig. 14 An example of a class activation map (CAM) [58]. A CNN network trained on ImageNet classified the left image as a “bridge pier”. A heatmap for the category of “bridge pier”, generated by a method called Grad-CAM [59], is superimposed (right image), which indicates the discriminative image regions used by the CNN for the classificationCC BY · PubMed Central
Fig. 15 An adversarial example demonstrated by Goodfellow et al. [61]. A network classified the object in the left image as a “panda” with 57.7% confidence. By adding a very small amount of carefully constructed noise (middle image), the network misclassified the object as a “gibbon” with 99.3% confidence on the right image without a visible change to a human. Reprinted with permission from “Explaining and harnessing adversarial examples” by Goodfellow et al. [61]
Fig. 15 An adversarial example demonstrated by Goodfellow et al. [61]. A network classified the object in the left image as a “panda” with 57.7% confidence. By adding a very small amount of carefully constructed noise (middle image), the network misclassified the object as a “gibbon” with 99.3% confidence on the right image without a visible change to a human. Reprinted with permission from “Explaining and harnessing adversarial examples” by Goodfellow et al. [61]CC BY · PubMed Central

Verification

Claims the paper makes

23 Valid · 11 Not valid · 6 Not checked · 47 Own results and interpretations

18 more claims · 8 with their cited sources checked

Create a free account to see all 87 claims and to verify the 2 references still open.

See all claims — free account
  • Probably not a substantial claim. 56
  • It may state two separate claims as one. 31

Create a free account to see these claims.

Citations

Cites

  • Deep Residual Learning for Image Recognition

    He, K., Zhang, X., Ren, S., & Sun, J. (2016) · 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2016.90 · Cited by 235,474 · cited 1×

    “At present, many models pretrained on the ImageNet challenge dataset are open to the public and readily accessible, along with their learned kernels and weights, such as AlexNet [3], VGG [30], ResNet [31], Inception [32], and DenseNet [33].”
    Method
  • Very Deep Convolutional Networks for Large-Scale Image Recognition

    Simonyan, K., & Zisserman, A. (2014) · arXiv · doi:10.48550/arxiv.1409.1556 · Cited by 114,148 · cited 2×

    “To perform unsupervised categorization, the following three steps were iteratively performed: (i) feature extraction using pretrained VGG16 model [30] from the lesions in the bounding box, (ii) clustering of the features, and (iii) fine-tuning of VGG16 based on the results of the clustering.”
    Method
  • U-Net: Convolutional Networks for Biomedical Image Segmentation

    Ronneberger, O., Fischer, P., & Brox, T. (2015) · Lecture notes in computer science · doi:10.1007/978-3-319-24574-4_28 · Cited by 99,404 · cited 1×

    “Although segmentation based on image patch was successfully performed in deep learning, U-net of Ronneberger et al. [48] outperformed the image patch-based method on the ISBI [IEEE (The Institute of Electrical and Electronics Engineers) International Symposium on Biomedical Imaging] challenge for…”
    Method
  • Adam: A Method for Stochastic Optimization

    Kingma, D. P., & Ba, J. (2014) · arXiv · doi:10.48550/arxiv.1412.6980 · Cited by 98,946

  • Deep learning

    LeCun, Y., Bengio, Y., & Hinton, G. (2015) · Nature · doi:10.1038/nature14539 · Cited by 84,126

  • Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks

    Ren, S., He, K., Girshick, R., & Sun, J. (2017) · IEEE Transactions on Pattern Analysis and Machine Intelligence · doi:10.1109/tpami.2016.2577031 · Cited by 73,472 · cited 1×

    “The study also suggested that the detection system was built using the Faster R-CNN method [56], the lesions in the bounding box, and their corresponding pseudo-category.”
    Method
  • Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks

    Ren, S., He, K., Girshick, R., & Sun, J. (2015) · arXiv · doi:10.48550/arxiv.1506.01497 · Cited by 70,812

  • Densely Connected Convolutional Networks

    Huang, G., Liu, Z., Van Der Maaten, L., & Weinberger, K. Q. (2017) · 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2017.243 · Cited by 48,566 · cited 1×

    “At present, many models pretrained on the ImageNet challenge dataset are open to the public and readily accessible, along with their learned kernels and weights, such as AlexNet [3], VGG [30], ResNet [31], Inception [32], and DenseNet [33].”
    Method
  • Going deeper with convolutions

    Szegedy, C., Liu, W., Jia, Y., Sermanet, P., Reed, S., Anguelov, D., Erhan, D., Vanhoucke, V., & Rabinovich, A. (2015) · 2015 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2015.7298594 · Cited by 48,562 · cited 5×

    “According to the results, the best area under the curve of receiver operating characteristic curves for detecting pulmonary tuberculosis from healthy cases was 0.99, which was obtained by ensemble of the AlexNet and GoogLeNet 2D-CNNs.”
    MethodInfluential
  • ImageNet Large Scale Visual Recognition Challenge

    Russakovsky, O., Deng, J., Su, H., Krause, J., Satheesh, S., Ma, S., Huang, Z., Karpathy, A., Khosla, A., Bernstein, M., et al. (2015) · International Journal of Computer Vision · doi:10.1007/s11263-015-0816-y · Cited by 39,380 · cited 1×

    “…neural network (CNN), a class of artificial neural networks that has been a dominant method in computer vision tasks since the astonishing results were shared on the object recognition competition known as the ImageNet Large Scale Visual Recognition Competition (ILSVRC) in 2012 [2, 3].”
    Method

Show all 55 references

Research works printed but not linked: 3

  1. Krizhevsky A, Sutskever I, Hinton GE (2012) ImageNet classification with deep convolutional neural networks. Adv Neural Inf Process Syst 25. Available online at: https://papers.nips.cc/paper/4824-imagenet-classification-with-deep-convolutional-neural-networks.pdf . Accessed 22 Jan 2018
  2. Lee DH (2013) Pseudo-label: the simple and efficient semi-supervised learning method for deep neural networks. In: Proceedings of the ICML 2013 Workshop: Challenges in Representation Learning. Available online at: http://deeplearning.net/wp-content/uploads/2013/03/pseudo_label_final.pdf . Accessed 23 Jan 2018
  3. Kurata Y, Nishio M, Fujimoto K et al (2018) Automatic segmentation of uterus with malignant tumor on MRI using U-net. In: Proceedings of the Computer Assisted Radiology and Surgery (CARS) 2018 congress (accepted)

Cited by

See all 4,682

Metadata

Authors
Yamashita, R., Nishio, M., Do, R. K. G., & Togashi, K.
Affiliations
  • Memorial Sloan Kettering Cancer Center (US)
  • Kyoto University (JP)
Corresponding author
Yamashita, Rikiya
Venue
Insights into Imaging
Published
2018-06-22
Volume
9
Issue
4
Pages
611-629
ISSN
1869-4101 (ISSN-L)
OpenAlex
W2809254203
Cited by
4,682
References
55
Citations by year
2026: 390 · 2025: 701 · 2024: 915 · 2023: 889 · 2022: 772 · 2021: 592 · 2020: 312 · 2019: 97 · 2018: 12
FWCI
5.66
Type
journal article
Language
en
Open access
Licence
Locations
  • Insights into Imaging (publishedVersion, cc-by)
  • Europe PMC (PubMed Central) (publishedVersion, other-oa)
  • PubMed Central (publishedVersion, cc-by)
  • PubMed (publishedVersion)
  • DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
Funders
Japan Society for the Promotion of Science (JP16K19883)
Topics
  • Health Sciences › Medicine › Health Informatics › Artificial Intelligence in Healthcare and Education
  • Health Sciences › Medicine › Radiology, Nuclear Medicine and Imaging › Radiomics and Machine Learning in Medical Imaging
  • Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Advanced Neural Network Applications
Keywords
convolutional neural networks, deep learning in radiology, convolutional layers, pooling layers, small datasets, overfitting, computer-aided diagnosis, radiology image analysis

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.

The claim verdicts this report counts are experimental: their accuracy is not yet measured.

A language model drew the claims from the paper on 2026-10-10, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).

A second model labelled each claim on 2026-10-10 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).

Cited sources are checked with a language model’s help (gpt-6-sol).

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

How to cite

Cite CoChat for the checks on this page, and the paper for its own findings.

This page
CoChat. (2026). Citation check: Convolutional neural networks: an overview and application in radiology. https://app.cochat.ai/papers/10.1007/s13244-018-0639-9
The paper
Yamashita, R., Nishio, M., Do, R. K. G., & Togashi, K. (2018). Convolutional neural networks: an overview and application in radiology. Insights into Imaging. https://doi.org/10.1007/s13244-018-0639-9

Data sources

Bibliographic data: Semantic Scholar, OpenAlex, Crossref.

Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).

Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.