Transfer learning classification of suspicious lesions on breast ultrasound: is there room to avoid biopsies of benign lesions?
Artificial intelligence
Breast neoplasms
Machine learning
Neural networks (computer)
Ultrasound
Journal
European radiology experimental
ISSN: 2509-9280
Titre abrégé: Eur Radiol Exp
Pays: England
ID NLM: 101721752
Informations de publication
Date de publication:
28 Oct 2024
28 Oct 2024
Historique:
received:
15
04
2024
accepted:
19
05
2024
medline:
28
10
2024
pubmed:
28
10
2024
entrez:
28
10
2024
Statut:
epublish
Résumé
Breast cancer (BC) is the most common malignancy in women and the second cause of cancer death. In recent years, there has been a strong development in artificial intelligence (AI) applications in medical imaging for several tasks. Our aim was to evaluate the potential of transfer learning with convolutional neural networks (CNNs) in discriminating suspicious breast lesions on ultrasound images. Transfer learning performances of five different CNNs (Inception V3, Xception, Densenet121, VGG 16, and ResNet50) were evaluated on a public and on an institutional dataset (526 and 392 images, respectively), customizing the top layers for the specific task. Institutional images were contoured by an expert radiologist and processed to feed the CNNs for training and testing. Postimaging biopsies were used as a reference standard for classification. The area under the receiver operating curve (AUROC) was used to assess diagnostic performance. Networks performed very well on the public dataset (AUROC 0.938-0.996). The direct generalization to the institutional dataset resulted in lower performances (max AUROC 0.676); however, when tested on BI-RADS 3 and BI-RADS 5 only, results were improved (max AUROC 0.792). Good results were achieved on the institutional dataset (AUROC 0.759-0.818) and, when selecting a threshold of 2% for classification, a sensitivity of 0.983 was obtained for three of five CNNs, with the potential to spare biopsy in 15.3%-18.6% of patients. In conclusion, transfer learning with CNNs may achieve high sensitivity and might be used as a support tool in managing suspicious breast lesions on ultrasound images. Transfer learning is a powerful technique to exploit the performances of well-trained CNNs for image classification. In a clinical scenario, it might be useful for the management of suspicious breast lesions on breast ultrasound, potentially sparing biopsy in a non-negligible number of patients. Properly trained CNNs with transfer learning are highly effective in differentiating benign and malignant lesions on breast ultrasound. Setting clinical thresholds increased sensitivity. CNNs might be useful as support tools in managing suspicious lesions on breast ultrasound.
Sections du résumé
BACKGROUND
BACKGROUND
Breast cancer (BC) is the most common malignancy in women and the second cause of cancer death. In recent years, there has been a strong development in artificial intelligence (AI) applications in medical imaging for several tasks. Our aim was to evaluate the potential of transfer learning with convolutional neural networks (CNNs) in discriminating suspicious breast lesions on ultrasound images.
METHODS
METHODS
Transfer learning performances of five different CNNs (Inception V3, Xception, Densenet121, VGG 16, and ResNet50) were evaluated on a public and on an institutional dataset (526 and 392 images, respectively), customizing the top layers for the specific task. Institutional images were contoured by an expert radiologist and processed to feed the CNNs for training and testing. Postimaging biopsies were used as a reference standard for classification. The area under the receiver operating curve (AUROC) was used to assess diagnostic performance.
RESULTS
RESULTS
Networks performed very well on the public dataset (AUROC 0.938-0.996). The direct generalization to the institutional dataset resulted in lower performances (max AUROC 0.676); however, when tested on BI-RADS 3 and BI-RADS 5 only, results were improved (max AUROC 0.792). Good results were achieved on the institutional dataset (AUROC 0.759-0.818) and, when selecting a threshold of 2% for classification, a sensitivity of 0.983 was obtained for three of five CNNs, with the potential to spare biopsy in 15.3%-18.6% of patients.
CONCLUSION
CONCLUSIONS
In conclusion, transfer learning with CNNs may achieve high sensitivity and might be used as a support tool in managing suspicious breast lesions on ultrasound images.
RELEVANCE STATEMENT
CONCLUSIONS
Transfer learning is a powerful technique to exploit the performances of well-trained CNNs for image classification. In a clinical scenario, it might be useful for the management of suspicious breast lesions on breast ultrasound, potentially sparing biopsy in a non-negligible number of patients.
KEY POINTS
CONCLUSIONS
Properly trained CNNs with transfer learning are highly effective in differentiating benign and malignant lesions on breast ultrasound. Setting clinical thresholds increased sensitivity. CNNs might be useful as support tools in managing suspicious lesions on breast ultrasound.
Identifiants
pubmed: 39466515
doi: 10.1186/s41747-024-00480-y
pii: 10.1186/s41747-024-00480-y
doi:
Types de publication
Journal Article
Langues
eng
Sous-ensembles de citation
IM
Pagination
121Informations de copyright
© 2024. The Author(s).
Références
Ferlay J, Steliarova-Foucher E, Lortet-Tieulent J et al (2013) Cancer incidence and mortality patterns in Europe: estimates for 40 countries in 2012. Eur J Cancer 49:1374–1403. https://doi.org/10.1016/j.ejca.2012.12.027
doi: 10.1016/j.ejca.2012.12.027
pubmed: 23485231
Yang L, Wang S, Zhang L et al (2020) Performance of ultrasonography screening for breast cancer: a systematic review and meta-analysis. BMC Cancer 20:1–15. https://doi.org/10.1186/s12885-020-06992-1
doi: 10.1186/s12885-020-06992-1
Drukker K, Giger ML, Horsch K, Kupinski MA, Vyborny CJ, Mendelson EB (2002) Computerized lesion detection on breast ultrasound. Med Phys 29:1438–1446. https://doi.org/10.1118/1.1485995
doi: 10.1118/1.1485995
pubmed: 12148724
Sickles EA, D’Orsi CJ, Bassett LW et al (2013) ACR BI-RADS mammography. In: ACR BI-RADS Atlas: breast imaging reporting and data system. American College of Radiology, Reston, pp 134–136. https://www.acr.org/Clinical-Resources/Reporting-and-Data-Systems/Bi-Rads/Permissions
Crystal P, Strano SD, Shcharynski S, Koretz MJ (2003) Using sonography to screen women with mammographically dense breasts. AJR Am J Roentgenol 181:177–182. https://doi.org/10.2214/ajr.181.1.1810177
doi: 10.2214/ajr.181.1.1810177
pubmed: 12818853
Lazarus E, Mainiero MB, Schepps B, Koelliker SL, Livingston LS (2006) BI-RADS lexicon for US and mammography: interobserver variability and positive predictive value. Radiology 239:385–391. https://doi.org/10.1148/radiol.2392042127
doi: 10.1148/radiol.2392042127
pubmed: 16569780
Litjens G, Kooi T, Bejnordi BE et al (2017) A survey on deep learning in medical image analysis. Med Image Anal 42:60–88. https://doi.org/10.1016/j.media.2017.07.005
doi: 10.1016/j.media.2017.07.005
pubmed: 28778026
He K, Zhang X, Ren S, Sun J (2016) Deep residual learning for image recognition. IEEE conference on computer vision and pattern recognition, IEEE, Las Vegas, 27–30 June 2016. https://doi.org/10.1109/CVPR.2016.90
Szegedy C, Vanhoucke V, Ioffe S, Shlens J, Wojna Z (2016) Rethinking the inception architecture for computer vision. IEEE conference on computer vision and pattern recognition, IEEE, Las Vegas, 27–30 June 2016. https://doi.org/10.1109/CVPR.2016.308
Chollet F (2017) Xception: deep learning with depthwise separable convolutions. IEEE conference on computer vision and pattern recognition, IEEE, Honolulu, 21–26 July 2017. https://doi.org/10.1109/CVPR.2017.195
Simonyan K, Zisserman A (2015) Very deep convolutional networks for large-scale image recognition. 3rd International Conference on Learning Representations (ICLR 2015), 1–14
Huang G, Liu Z, Van der Maaten L, Weinberger KQ (2017) Densely connected convolutional networks. IEEE conference on computer vision and pattern recognition, IEEE, Honolulu, 21–26 July 2017. https://doi.org/10.1109/CVPR.2017.243
Yosinski J, Clune J, Bengio Y, Lipson H (2014). How transferable are features in deep neural networks? ArXiv, abs/1411.1792
Ayana G, Dese K, Choe S (2021) Transfer learning in breast cancer diagnoses via ultrasound imaging. Cancers 13:738. https://doi.org/10.3390/cancers13040738
doi: 10.3390/cancers13040738
pubmed: 33578891
pmcid: 7916666
Ali MD, Saleem A, Elahi H et al (2023) Breast cancer classification through meta-learning ensemble technique using convolution neural networks. Diagnostics (Basel) 13:2242. https://doi.org/10.3390/diagnostics13132242
doi: 10.3390/diagnostics13132242
pubmed: 37443636
Xiao T, Liu L, Kai L, Wenjian Q, Shaode Y, Zhicheng L (2018) Comparison of transferred deep neural networks in ultrasonic breast masses discrimination. Biomed Res Int 4605191. https://doi.org/10.1155/2018/4605191
Jafari Z, Karami E (2023) Breast cancer detection in mammography images: a CNN-based approach with feature selection. Information 14:410. https://doi.org/10.3390/info14070410
doi: 10.3390/info14070410
Li J, Bu Y, Lu S et al (2021) Development of a deep learning-based model for diagnosing breast nodules with ultrasound. J Ultrasound Med 40:513–520. https://doi.org/10.1002/jum.15427
doi: 10.1002/jum.15427
pubmed: 32770574
Hijab A, Rushdi MA, Gomaa MM, Eldeib A (2019) Breast cancer classification in ultrasound images using transfer learning. Fifth International Conference on Advances in Biomedical Engineering, IEEE, Lebanon, 17–19 October 2019. https://doi.org/10.1109/ICABME47164.2019.8940291
Al-Dhabyani W, Gomaa M, Khaled H, Fahmy A (2020) Dataset of breast ultrasound images. Data Brief 28:104863. https://doi.org/10.1016/j.dib.2019.104863
doi: 10.1016/j.dib.2019.104863
pubmed: 31867417
Deng J, Dong W, Socher R et al (2009) Imagenet: a large-scale hierarchical image database. IEEE conference on computer vision and pattern recognition, IEEE, Miami, 20–25 June 2009. https://doi.org/10.1109/CVPR.2009.5206848
Gulli A, Pal S Deep learning with Keras (2017) Packt Publishing Ltd
Abadi M, Agarwal A, Barham P et al (2015) TensorFlow: large-scale machine learning on heterogeneous systems. Software available from tensorflow.org
Pathan RK, Alam FI, Yasmin S et al (2022) Breast cancer classification by using multi-headed convolutional neural network modelling. Healthcare (Basel) 10:2367. https://doi.org/10.3390/healthcare10122367
doi: 10.3390/healthcare10122367
pubmed: 36553891
Zhang N, Li XT, Ma L, Fan ZQ, Sun YS (2021) Application of deep learning to establish a diagnostic model of breast lesions using two-dimensional grayscale ultrasound imaging. Clin Imaging 79:56–63. https://doi.org/10.1016/j.clinimag.2021.03.024
doi: 10.1016/j.clinimag.2021.03.024
pubmed: 33887507
Fujioka T, Kubota K, Mori M et al (2019) Distinction between benign and malignant breast masses at breast ultrasound using deep learning method with convolutional neural network. Jpn J Radiol 37:466–472. https://doi.org/10.1007/s11604-019-00831-5
doi: 10.1007/s11604-019-00831-5
pubmed: 30888570
Gu Y, Xu W, Lin B et al (2022) Deep learning based on ultrasound images assists breast lesion diagnosis in China: a multicenter diagnostic study. Insights Imaging 13:124. https://doi.org/10.1186/s13244-022-01259-8
doi: 10.1186/s13244-022-01259-8
pubmed: 35900608
pmcid: 9334487
Polat DS, Merchant K, Hayes J, Omar L, Compton L, Dogan BS (2023) Outcome of imaging and biopsy of BI-RADS category 3 lesions: follow-Up compliance, biopsy, and malignancy rates in a large patient cohort. J Ultrasound Med 42:1285–1296. https://doi.org/10.1002/jum.16142
doi: 10.1002/jum.16142
pubmed: 36445017
Lee CS, Berg JM, Berg WA (2021) Cancer yield exceeds 2% for BI-RADS 3 probably benign findings in women older than 60 years in the national mammography database. Radiology 299:550–558. https://doi.org/10.1148/radiol.2021204031
doi: 10.1148/radiol.2021204031
pubmed: 33787333
Lyu SY, Zhang Y, Zhang MW et al (2022) Diagnostic value of artificial intelligence automatic detection systems for breast BI-RADS 4 nodules. World J Clin Cases 10:518–527. https://doi.org/10.12998/wjcc.v10.i2.518
doi: 10.12998/wjcc.v10.i2.518
pubmed: 35097077
pmcid: 8771370
Shen Y, Shamout FE, Oliver JR et al (2021) Artificial intelligence system reduces false-positive findings in the interpretation of breast ultrasound exams. Nat Commun 12:5645. https://doi.org/10.1038/s41467-021-26023-2
doi: 10.1038/s41467-021-26023-2
pubmed: 34561440
pmcid: 8463596
Hayashida T, Odani E, Kikuchi M et al (2022) Establishment of a deep‐learning system to diagnose BI‐RADS4a or higher using breast ultrasound for clinical application. Cancer Sci 113:3528–3534. https://doi.org/10.1111/cas.15511
doi: 10.1111/cas.15511
pubmed: 35880248
pmcid: 9530860
Qi X, Zhang L, Chen Y et al (2019) Automated diagnosis of breast ultrasonography images using deep neural networks. Med Image Anal 52:185–198. https://doi.org/10.1016/j.media.2018.12.006
doi: 10.1016/j.media.2018.12.006
pubmed: 30594771
Du R, Chen Y, Li T, Shi L, Fei ZD, Li Y (2022) Discrimination of breast cancer based on ultrasound images and convolutional neural network. J Oncol. https://doi.org/10.1155/2022/7733583
Hassanien MA, Singh VK, Puig D, Abdel-Nasser M (2022) Predicting breast tumor malignancy using deep ConvNeXt radiomics and quality-based score pooling in ultrasound sequences. Diagnostics (Basel) 12:1053. https://doi.org/10.3390/diagnostics12051053
doi: 10.3390/diagnostics12051053
pubmed: 35626208
Kumar V, Webb JM, Gregory A et al (2018) Automated and real-time segmentation of suspicious breast masses using convolutional neural network. PLoS One 13:e0195816. https://doi.org/10.1371/journal.pone.0195816
doi: 10.1371/journal.pone.0195816
pubmed: 29768415
pmcid: 5955504
Mújica-Vargas D, Matuz-Cruz M, García-Aquino C, Ramos-Palencia C (2022) Efficient system for delimitation of benign and malignant breast masses. Entropy 24:1775. https://doi.org/10.3390/e24121775
doi: 10.3390/e24121775
pubmed: 36554180
pmcid: 9777637
Zhang S, Liao M, Wang J et al (2023) Fully automatic tumor segmentation of breast ultrasound images with deep learning. J Appl Clin Med Phys 24:e13863. https://doi.org/10.1002/acm2.13863
doi: 10.1002/acm2.13863
pubmed: 36495018
Gao Y, Liu B, Zhu Y et al (2021) Detection and recognition of ultrasound breast nodules based on semi-supervised deep learning: a powerful alternative strategy. Quant Image Med Surg 11:2265–2278. https://doi.org/10.21037/qims-20-12b
doi: 10.21037/qims-20-12b
Marrón-Esquivel JM, Duran-Lopez L, Linares-Barranco A, Dominguez-Morales JP (2023) A comparative study of the inter-observer variability on Gleason grading against deep learning-based approaches for prostate cancer. Comput Biol Med 159:106856. https://doi.org/10.1016/j.compbiomed.2023.106856
doi: 10.1016/j.compbiomed.2023.106856
pubmed: 37075600