Are Convolutional Neural Networks Trained on ImageNet Images Wearing Rose-Colored Glasses?: A Quantitative Comparison of ImageNet, Computed Tomographic, Magnetic Resonance, Chest X-Ray, and Point-of-Care Ultrasound Images for Quality.
artificial intelligence
computed tomography
deep learning
image quality
magnetic resonance
point-of-care ultrasound
Journal
Journal of ultrasound in medicine : official journal of the American Institute of Ultrasound in Medicine
ISSN: 1550-9613
Titre abrégé: J Ultrasound Med
Pays: England
ID NLM: 8211547
Informations de publication
Date de publication:
Feb 2021
Feb 2021
Historique:
received:
02
06
2020
revised:
16
06
2020
accepted:
22
06
2020
pubmed:
7
8
2020
medline:
15
5
2021
entrez:
7
8
2020
Statut:
ppublish
Résumé
Deep learning for medical imaging analysis uses convolutional neural networks pretrained on ImageNet (Stanford Vision Lab, Stanford, CA). Little is known about how such color- and scene-rich standard training images compare quantitatively to medical images. We sought to quantitatively compare ImageNet images to point-of-care ultrasound (POCUS), computed tomographic (CT), magnetic resonance (MR), and chest x-ray (CXR) images. Using a quantitative image quality assessment technique (Blind/Referenceless Image Spatial Quality Evaluator), we compared images based on pixel complexity, relationships, variation, and distinguishing features. We compared 5500 ImageNet images to 2700 CXR, 2300 CT, 1800 MR, and 18,000 POCUS images. Image quality results ranged from 0 to 100 (worst). A 1-way analysis of variance was performed, and the standardized mean-difference effect size value (d) was calculated. ImageNet images showed the best image quality rating of 21.7 (95% confidence interval [CI], 0.41) except for CXR at 13.2 (95% CI, 0.28), followed by CT at 35.1 (95% CI, 0.79), MR at 31.6 (95% CI, 0.75), and POCUS at 56.6 (95% CI, 0.21). The differences between ImageNet and all of the medical images were statistically significant (P ≤ .000001). The greatest difference in image quality was between ImageNet and POCUS (d = 2.38). Point-of-care ultrasound (US) quality is significantly different from that of ImageNet and other medical images. This brings considerable implications for convolutional neural network training with medical images for various applications, which may be even more significant in the case of US images. Ultrasound deep learning developers should consider pretraining networks from scratch on US images, as training techniques used for CT, CXR, and MR images may not apply to US.
Types de publication
Journal Article
Langues
eng
Sous-ensembles de citation
IM
Pagination
377-383Informations de copyright
© 2020 American Institute of Ultrasound in Medicine.
Références
Xu J, Xue K, Zhang K. Current status and future trends of clinical diagnoses via image-based deep learning. Theranostics 2019; 9:7556-7565.
Wang S, Yang DM, Rong R, et al. Artificial intelligence in lung cancer pathology image analysis. Cancers (Basel) 2019; 11:1673.
Zhang Z, Sejdić E. Radiological images and machine learning: trends, perspectives, and prospects. Comput Biol Med 2019; 108:354-370.
Komura D, Ishikawa S. Machine learning approaches for pathologic diagnosis. Virchows Arch 2019; 475:131-138.
Dodge S, Karam L. Understanding how image quality affects deep neural networks. arXiv website. https://arxiv.org/pdf/1604.04004.pdf. Published April 21, 2016.
US Food and Drug Administration. Class 2 device recall Vscan extend. US Food and Drug Administration website. https://www.accessdata.fda.gov/scripts/cdrh/cfdocs/cfRES/res.cfm?id=173162. Updated July 16, 2020.
Miotto R, Wang F, Wang S, Jiang X, Dudley JT. Deep learning for healthcare: review, opportunities and challenges. Brief Bioinform 2018; 19:1236-1246.
Blaivas M, Blaivas L. Are all deep learning architectures alike for point-of-care ultrasound? Evidence from a cardiac image classification model suggests otherwise. J Ultrasound Med 2020; 39:1187-1194.
Shrimali KR. Image quality assessment: BRISQUE. Learn OpenCV website. https://www.learnopencv.com/image-quality-assessment-brisque/. Published June 20, 2018.
Chilamkurthy S, Ghosh R, Tanamala S, et al. Deep learning algorithms for detection of critical findings in head CT scans: a retrospective study. Lancet 2018; 392:2388-2396.
Singh R, Kalra MK, Nitiwarangkul C, et al. Deep learning in chest radiography: detection of findings and presence of change. PLoS One 2018; 13:e0204155.
Cha MJ, Chung MJ, Lee JH, Lee KS. Performance of deep learning model in detecting operable lung cancer with chest radiographs. J Thorac Imaging 2019; 34:86-91.
Ginat DT. Analysis of head CT scans flagged by deep learning software for acute intracranial hemorrhage. Neuroradiology 2020; 62:335-340.
Wiljer D, Hakim Z. Developing an artificial intelligence-enabled health care practice: rewiring health care professions for better care. J Med Imaging Radiat Sci 2019; 50(suppl 2):S8-S14.
Kaiming H, Xiangyu Z, Shaoqing R, Jian S. Deep residual learning for image recognition. In: Proceedings of the 2016 IEEE Conference on Computer Vision and Pattern Recognition. Piscataway, NJ: Institute of Electrical and Electronics Engineers; 2016:770-778.
Rajagopal H, Mokhtar N, Tengku Mohmed Noor Izam TF, Wan Ahmad WK. No-reference quality assessment for image-based assessment of economically important tropical woods. PLoS One 2020; 15:e0233320.
Mittal A, Moorthy A, Bovik A. No-reference image quality assessment in the spatial domain. IEEE Trans Image Process 2012; 21:4695-4708.
Zago GT, Andreão RV, Dorizzi B, Teatini Salles EO. Retinal image quality assessment using deep learning. Comput Biol Med 2018; 103:64-70.