Skip to main content
Log in

Study on diagnosis of thyroid nodules based on convolutional neural network

Studie zur Diagnose von Schilddrüsenknoten auf der Grundlage künstlicher neuronaler Netze

  • Original articles
  • Published:
Die Radiologie Aims and scope Submit manuscript

Abstract

Objective

An artificial intelligence (AI) algorithm based on convolutional neural networks was used in ultrasound diagnosis in order to evaluate its performance in judging the nature of thyroid nodules and nodule classification.

Methods

A total of 105 patients with thyroid nodules confirmed by surgery or biopsy were retrospectively analyzed. The properties, characteristics, and classification of thyroid nodules were evaluated by sonographers and by AI to obtain combined diagnoses. Receiver operating characteristic curves were generated to evaluate the performance of AI, the sonographer, and their combined effort in diagnosing the nature of thyroid nodules and classifying their characteristics. In the diagnosis of thyroid nodules with solid components, hypoechoic appearance, indistinct borders, Anteroposterior/transverse diameter ratio > 1(A/T > 1), and calcification performed by sonographers and by AI, the properties exhibited statistically significant differences.

Results

Sonographers had a sensitivity of 80.7%, specificity of 73.7%, accuracy of 79.0%, and area under the curve (AUC) of 0.751 in the diagnosis of benign and malignant thyroid nodules. AI had a sensitivity of 84.5%, specificity of 81.0%, accuracy of 84.7%, and AUC of 0.803. The combined AI and sonographer diagnosis had a sensitivity of 92.1%, specificity of 86.3%, accuracy of 91.7%, and AUC of 0.910.

Conclusion

The efficacy of a combined diagnosis for benign and malignant thyroid nodules is higher than that of an AI-based diagnosis alone or a sonographer-based diagnosis alone. The combined diagnosis can reduce unnecessary fine-needle aspiration biopsy procedures and better evaluate the necessity of surgery in clinical practice.

Zusammenfassung

Ziel

Ein Algorithmus der künstlichen Intelligenz (KI) auf der Basis künstlicher (gefalteter) neuronaler Netze wurde in der Ultraschalldiagnostik eingesetzt, um seine Leistungsfähigkeit bei der Beurteilung von Schilddrüsenknoten und der Knotenklassifikation zu untersuchen.

Methoden

Retrospektiv wurden die Daten von 105 Patienten mit chirurgisch oder bioptisch bestätigten Schilddrüsenknoten analysiert. Eigenschaften, Merkmale und Klassifikation der Schilddrüsenknoten wurden von den Ultraschalluntersuchern sowie durch KI beurteilt, um so kombinierte Diagnosen zu erhalten. Mithilfe von ROC(„receiver operating characteristic curves“)-Kurven wurde die Leistungsfähigkeit der KI, der Ultraschalluntersucher und des kombinierten Einsatzes beider ermittelt, die Art der Schilddrüsenknoten und ihrer Klassifizierung zu beurteilen. Bei der durch Ultraschalluntersucher und KI erfolgten Diagnose von Schilddrüsenknoten mit festen Anteilen, echoarmem Erscheinungsbild, unklaren Grenzen, Vorne und hinten/Seitendurchmesserverhältnis > 1 sowie Kalzifizierung wiesen die Eigenschaften statistisch signifikante Unterschiede auf.

Ergebnisse

Bei den Ultraschalluntersuchern betrug die Sensitivität 80,7%, die Spezifität 73,7%, die Genauigkeit 79,0% und die AUC („area under the curve“) 0,751 für die Diagnose benigner und maligner Schilddrüsenknoten. Für die KI lag die Sensitivität bei 84,5%, die Spezifität bei 81,0%, die Genauigkeit bei 84,7% und die AUC bei 0,803. Die Diagnose auf der Grundlage einer Kombination aus KI und Ultraschalluntersuchern wies eine Sensitivität von 92,1%, ein Spezifität von 86,3%, eine Genauigkeit von 91,7% und eine AUC von 0,910 auf.

Schlussfolgerung

Die Verlässlichkeit einer Diagnose benigner und maligner Schilddrüsenknoten auf der Basis einer Kombination aus KI und Ultraschalluntersuchern ist höher als die einer KI-basierten Diagnose allein oder einer alleinigen Diagnose durch Ultraschalluntersucher. Mit einer Diagnose aus der Kombination beider Ansätze können unnötige Feinnadelaspirationsbiopsien vermindert und die Notwendigkeit eines chirurgischen Eingriffs im klinischen Alltag besser beurteilt werden.

This is a preview of subscription content, log in via an institution to check access.

Access this article

Price excludes VAT (USA)
Tax calculation will be finalised during checkout.

Instant access to the full article PDF.

Fig. 1
Fig. 2
Fig. 3
Fig. 4
Fig. 5

Similar content being viewed by others

References

  1. Kim TY, Shong YK (2017) Active surveillance of papillary thyroid microcarcinoma: a mini-review from Korea. Endocrinol Metab 32(4):399–406

    Article  Google Scholar 

  2. Zahir ST, Vakili M, Ghaneei A et al (2016) Ultrasound assistance in differentiating malignant thyroid nodules from benign ones. J Ayub Med Coll Abbottabad 28(4):644–649

    PubMed  Google Scholar 

  3. Zhang Y, Zhou P, Tian SM et al (2017) Usefulness of combined use of contrast-enhanced ultrasound and TI-RADS classification for the differe ntiation of benign from malignant lesions of thyroid nodules. Eur Radiol 27(4):1527–1536

    Article  PubMed  Google Scholar 

  4. Mciver B, Hay ID, Giuffrida DF et al (2001) Anaplastic thyroid carcinoma: a 50-year experience at a single institution. Surgery 130(6):1028–1034

    Article  CAS  PubMed  Google Scholar 

  5. Liang XW, Cai YY, Yu JS et al (2019) Update on thyroid ultrasound: a narrative review from diagnostic criteria to artificial intelligence techniques. Chin Med J 132(16):1974–1982

    Article  PubMed  PubMed Central  Google Scholar 

  6. Tessler FN, Middleton WD, Grant EG et al (2017) ACR Thyroid Imaging, Reporting and Data System (TI-RADS): white paper of the ACR TI-RADS committee. J Am Coll Radiol 14(5):587–595

    Article  PubMed  Google Scholar 

  7. Kermany DS, Goldbaum M, Cai W et al (2018) Identifying medical diagnoses and treatable diseases by image-based deep learning. Cell 172(5):1122–1131.e9

    Article  CAS  PubMed  Google Scholar 

  8. Li X, Zhang S, Zhang Q et al (2019) Diagnosis of thyroid cancer using deep convolutional neural network models applied to sonographic ima ges: a retrospective, multicohort, diagnostic study. Lancet Oncol 20(2):193–201

    Article  PubMed  Google Scholar 

  9. Song W, Li S, Liu J et al (2019) Multitask cascade convolution neural networks for automatic thyroid nodule detection and recognition. IEEE J Biomed Health Inform 23(3):1215–1224

    Article  PubMed  Google Scholar 

  10. Choi YJ, Baek JH, Park HS et al (2017) A computer-aided diagnosis system using artificial intelligence for the diagnosis and characterization of thyroid nodules on ultrasound: initial clinical assessment. Thyroid 27(4):546–552

    Article  PubMed  Google Scholar 

  11. Wang S, Xu J, Tahmasebi A et al (2020) Incorporation of a machine learning algorithm with object detection within the thyroid imaging reporting and data system improves the diagnosis of genetic risk. Front Oncol 10:591846

    Article  PubMed  PubMed Central  Google Scholar 

  12. Thomas J, Haertling T (2020) AIBx, Artificial Intelligence model to risk stratify thyroid nodules. Thyroid 30(6):878–884

    Article  PubMed  Google Scholar 

  13. Wei X, Zhu J, Zhang H et al (2020) Visual interpretability in computer-assisted diagnosis of thyroid nodules using ultrasound images. Med Sci Monit 26:e927007

    Article  PubMed  PubMed Central  Google Scholar 

  14. Wang J, Jiang J, Zhang D et al (2022) An integrated AI model to improve diagnostic accuracy of ultrasound and output known risk features in suspicious thyroid nodules. Eur Radiol 32(3):2120–2129

    Article  PubMed  Google Scholar 

  15. Zhang B, Tian J, Pei S et al (2019) Machine learning-assisted system for thyroid nodule diagnosis. Thyroid 29(6):858–867

    Article  PubMed  Google Scholar 

  16. Wei X, Gao M, Yu R et al (2020) Ensemble deep learning model for multicenter classification of thyroid nodules on ultrasound images. Med Sci Monit 26:e926096

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  17. Chambara N, Ying M (2019) The diagnostic efficiency of ultrasound computer-aided diagnosis in differentiating thyroid nodules: a systematic review and narrative synthesis. Cancers 11(11):1759

    Article  PubMed  PubMed Central  Google Scholar 

  18. Zhang T, Li F, Mu J, Liu J, al Zhet (2017) Multivariate evaluation of Thyroid Imaging Reporting and Data System (TI-RADS) in diagnosis malignant thyroid nodule: application to PCA and PLS-DA analysis. Int J Clin Oncol 22(3):448–454

    Article  PubMed  Google Scholar 

  19. Gong T, Wang J (2012) The analysis of the calcification in differentiating malignant thyroid neoplasm and the molecular me chanisms for the formation of the calcification. Lin Chung Er Bi Yan Hou Tou Jing Wai Ke Za Zhi 26(16):763–766

    CAS  Google Scholar 

  20. Frates MC, Benson CB, Charboneau JW et al (2006) Management of thyroid nodules detected at US: Society of Radiologists in Ultrasound consensus confere nce statement. Ultrasound Q 22(4):231–238 (discussion 9–40)

    Article  PubMed  Google Scholar 

  21. Su JJ, Hui LZ, Xi CJ, Su GQ (2015) Correlation analysis of ultrasonic characteristics, pathological type, and molecular markers of thyroid nodules. Genet Mol Res 14(1):9–20

    Article  CAS  PubMed  Google Scholar 

  22. Moon HJ, Kwak JY, Kim EK, Kim MJ (2011) A taller-than-wide shape in thyroid nodules in transverse and longitudinal ultrasonographic planes and the prediction of malignancy. Thyroid 21(11):1249–1253

    Article  PubMed  Google Scholar 

  23. Zhang S, Zhao J, Xin XJ et al (2013) Diagnostic value of thyroid microcarcinoma with a taller-than-wide shape in thyroid nodules. Zhonghua Yi Xue Za Zhi 93(40):3223–3225

    PubMed  Google Scholar 

  24. Desser TS, Kamaya A (2008) Ultrasound of thyroid nodules. Neuroimaging Clin N Am 18(3):463–478

    Article  PubMed  Google Scholar 

  25. Yao S, Yan J, Wu M et al (2020) Texture synthesis based thyroid nodule detection from medical ultrasound images: interpreting and sup pressing the adversarial effect of in-place manual annotation. Front Bioeng Biotechnol 8:599

    Article  PubMed  PubMed Central  Google Scholar 

  26. Akkus Z, Cai J, Boonrod A et al (2019) A survey of deep-learning applications in ultrasound: artificial intelligence-powered ultrasound for improving clinical workflow. J Am Coll Radiol 16(9 Pt B):1318–1328

    Article  PubMed  Google Scholar 

Download references

Funding

Yunnan Province “Ten thousand people plan” famous doctor special (certificate number: YNWR-MY-2018-004).

Author information

Authors and Affiliations

Authors

Contributions

Aitao Yin and Yongping Lu designed the research study, Fei Xu analyzed the data, Yifan Zhao, Yue Sun, Miao Huang, and Xiangbi Li wrote the paper.

Corresponding author

Correspondence to YongPing Lu.

Ethics declarations

Conflict of interest

A. Yin, Y. Lu, F. Xu, Y. Zhao, Y. Sun, M. Huang and X. Li declare that they have no competing interests.

All studies mentioned were in accordance with the ethical standards indicated in each case. This retrospective study was performed after consultation with the institutional ethics committee (Institutional Review Committee of the Affiliated Hospital of Yunnan University (approval number: 2021082)) and in accordance with national legal requirements. All patients were exempted from providing informed consent.

The supplement containing this article is not sponsored by industry.

Additional information

The authors AiTao Yin and Fei Xu contributed equally to this study.

figure qr

Scan QR code & read article online

Rights and permissions

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Yin, A., Lu, Y., Xu, F. et al. Study on diagnosis of thyroid nodules based on convolutional neural network. Radiologie 63 (Suppl 2), 64–72 (2023). https://doi.org/10.1007/s00117-023-01137-4

Download citation

  • Accepted:

  • Published:

  • Issue Date:

  • DOI: https://doi.org/10.1007/s00117-023-01137-4

Keywords

Schlüsselwörter

Navigation