Skip to main page content
U.S. flag

An official website of the United States government

Dot gov

The .gov means it’s official.
Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

Https

The site is secure.
The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
. 2023 Apr 20:17:1150120.
doi: 10.3389/fnhum.2023.1150120. eCollection 2023.

Pre-trained deep learning models for brain MRI image classification

Srigiri Krishnapriya  1 , Yepuganti Karuna  1
Affiliations

Pre-trained deep learning models for brain MRI image classification

Srigiri Krishnapriya et al. Front Hum Neurosci. .

Abstract

Brain tumors are serious conditions caused by uncontrolled and abnormal cell division. Tumors can have devastating implications if not accurately and promptly detected. Magnetic resonance imaging (MRI) is one of the methods frequently used to detect brain tumors owing to its excellent resolution. In the past few decades, substantial research has been conducted in the field of classifying brain images, ranging from traditional methods to deep-learning techniques such as convolutional neural networks (CNN). To accomplish classification, machine-learning methods require manually created features. In contrast, CNN achieves classification by extracting visual features from unprocessed images. The size of the training dataset had a significant impact on the features that CNN extracts. The CNN tends to overfit when its size is small. Deep CNNs (DCNN) with transfer learning have therefore been developed. The aim of this work was to investigate the brain MR image categorization potential of pre-trained DCNN VGG-19, VGG-16, ResNet50, and Inception V3 models using data augmentation and transfer learning techniques. Validation of the test set utilizing accuracy, recall, Precision, and F1 score showed that the pre-trained VGG-19 model with transfer learning exhibited the best performance. In addition, these methods offer an end-to-end classification of raw images without the need for manual attribute extraction.

Keywords: ResNet50; VGG-16; VGG-19; convolutional neural networks; inception V3; transfer learning.

PubMed Disclaimer

Conflict of interest statement

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Figures

Figure 1
Figure 1
Flowchart of the proposed methodology using pre-trained DCCN models.
Figure 2
Figure 2
Pie chart showing the distribution of tumorous and non-tumorous images in the dataset.
Figure 3
Figure 3
MR images with and without tumor from the dataset.
Figure 4
Figure 4
Steps involved in preprocessing.
Figure 5
Figure 5
Basic CNN architecture (Hinton, 2012).
Figure 6
Figure 6
Illustration of the network architecture of VGG-19 model (Abuared et al., 2020).
Figure 7
Figure 7
The standard VGG-16 network architecture (Simonyan and Zisserman, 2014).
Figure 8
Figure 8
The architecture of ResNet50 model (Khan et al., 2021).
Figure 9
Figure 9
Architecture of inception V3 model (Szegedy et al., 2016).
Figure 10
Figure 10
Accuracy and loss graphs for the VGG-19 model.
Figure 11
Figure 11
Accuracy and loss graphs for the VGG-16 model.
Figure 12
Figure 12
Accuracy and loss graphs of the ResNet50 model.
Figure 13
Figure 13
Accuracy and loss graphs of Inception V3 model.
Figure 14
Figure 14
The confusion matrix for classification models (A) Inception V3 (B) ResNet50 (C) VGG-16 (D) VGG-19.
Figure 15
Figure 15
Evaluation metric comparison of various classification models.
Figure 16
Figure 16
Comparison of model accuracies with and without augmentation.
See this image and copyright information in PMC

References

    1. Abiwinanda N., Hanif M., Hesaputra S. T., Handayani A., Mengko T. R. (2019). Brain Tumor Classification Using Convolutional Neural Network.” 183–89 in World congress on medical physics and biomedical engineering 2018. Berlin: Springer. 10.1007/978-981-10-9035-6_33 - DOI
    1. Abuared N., Panthakkan A., Al-Saad M., Amin S. A., Mansoor W. (2020). “Skin cancer classification model based on VGG 19 and transfer learning,” in 2020 3rd International Conference on Signal Processing and Information Security (ICSPIS) (IEEE: ). 10.1109/ICSPIS51252.2020.9340143 - DOI
    1. Afshar P., Mohammadi A., Plataniotis K. N. (2018). “Brain tumor type classification via Capsule Networks,” in 3129–3133 in 2018 25th IEEE International Conference on Image Processing (ICIP) (IEEE: ). 10.1109/ICIP.2018.8451379 - DOI
    1. Agn M., af Rosenschöld P. M., Puonti O., Lundemann M. J., Mancini L., Papadaki A., et al. . (2019). A modality-adaptive method for segmenting brain tumors and organs-at-risk in radiation therapy planning. Med. Image Anal. 54, 220–237. 10.1016/j.media.2019.03.005 - DOI - PMC - PubMed
    1. Ahmad S., Choudhury P. K. (2022). On the performance of deep transfer learning networks for brain tumor detection using MR images. IEEE Access. 10, 59099–59114. 10.1109/ACCESS.2022.3179376 - DOI