Investigation of neural network models application in EU-TIRADS thyroid nodules classification for personalization of thyroid gland ultrasound diagnostic

SUBSTANTIATION. It is known that about 69% of all thyroid nodules undergoing surgical treatment are benign formations, and up to 75% of patients with an intermediate cytological conclusion undergo unnecessary surgical intervention. This suggests that improving the quality of differential diagnosis of nodular formations will avoid excessive economic costs for the healthcare system. In this regard, AI technologies in diagnostic algorithms for the classification of thyroid nodules were involved.
AIM. Improving the efficiency of automatic classification of thyroid nodules on ultrasound images by using a set of neural network models.
MATERIALS AND METHODS. We used ultrasound images of thyroid nodules available in open sources and obtained with the help of 3 ultrasound devices of Endocrinology Research Centre as part of Project № 22-15-00135 of the grant of the Russian Science Foundation. This article check the hypothesis that the size of the training set cannot be increased by repeating similar images from the ultrasound cine loop of one patient, but only by expanding the dataset with new unique specimens of other patients and/or data from the augmentation process.
RESULTS. As a result, a neural network model EfficientNet-B6 was proposed to solve the problem of EU-TIRADS classification of thyroid nodules based on ultrasound images of the thyroid gland.
CONCLUSION. The results obtained allow us to advance in the use of artificial intelligence methods for personalized medicine in thyroid diseases.

1. Brockwell PJ, Davis RA. Introduction to time series and forecasting. New York: Springer; 2016.

2. Aldweesh A, Derhab A, Emam AZ. Deep learning approaches for anomaly-based intrusion detection systems: A survey, taxonomy, and open issues. Knowledge-Based Syst. 2020;(189):105124. doi: https://doi.org/10.1016/j.knosys.2019.105124

3. Chen J, You H, Li K. A review of thyroid gland segmentation and thyroid nodule segmentation methods for medical ultrasound images. Comput Methods Programs Biomed. 2020;(185):105329. doi: https://doi.org/10.1016/j.cmpb.2020.105329

4. Deng P, Han X, Wei X, Chang L. Automatic classification of thyroid nodules in ultrasound images using a multi-task attention network guided by clinical knowledge. Comput Biol Med. 2022;(150):106172. doi: https://doi.org/10.1016/j.compbiomed.2022.106172

5. Zhang X, Lee VC, Rong J, et al. Deep convolutional neural networks in thyroid disease detection: A multi-classification comparison by ultrasonography and computed tomography. Comput Methods Programs Biomed. 2022;(220):106823. doi: https://doi.org/10.1016/j.cmpb.2022.106823

6. Sharifi Y, Bakhshali MA, Dehghani T, et al. Deep learning on ultrasound images of thyroid nodules. Biocybern Biomed Eng. 2021;41(2):636-655. doi: https://doi.org/10.1016/j.bbe.2021.02.008

7. Tessler FN, Middleton WD, Grant EG. Thyroid Imaging Reporting and Data System (TI-RADS): A user’s guide. Radiology. 2018;287(1):29-36. doi: https://doi.org/10.1148/radiol.2017171240

8. Redmon J, Divvala S, Girshick R, et al. You only look once: unified, realtime object detection. Nucl Instruments Methods Phys Res Sect A Accel Spectrometers, Detect Assoc Equip. 2015 [cited 15.08.2023]. Available from: http://arxiv.org/abs/1506.02640

9. Chen LC, Papandreou G, Kokkinos I, et al. Rethinking Atrous Convolution for Semantic Image Segmentation Liang-Chieh. IEEE Trans Pattern Anal Mach Intell. 2018 [cited 15.08.2023]. Available from: http://arxiv.org/abs/1706.05587

10. Maharana K, Mondal S, Nemade B. A review: Data pre-processing and data augmentation techniques. Glob Transitions Proc. 2022;3(1):91-99. doi: https://doi.org/10.1016/j.gltp.2022.04.020

11. Nalepa J, Marcinkiewicz M, Kawulok M. Data augmentation for braintumor segmentation: A review. Front Comput Neurosci. 2019;(13). doi: https://doi.org/10.3389/fncom.2019.00083

12. Chlap P, Min H, Vandenberg N, et al. A review of medical image data augmentation techniques for deep learning applications. J Med Imaging Radiat Oncol. 2021;65(5):545-563. doi: https://doi.org/10.1111/1754-9485.13261

13. Hoar D, Lee PQ, Guida A, et al. Combined transfer learning and testtime augmentation improves convolutional neural network-based semantic segmentation of prostate cancer from multi-parametric MR images. Comput Methods Programs Biomed. 2021;(210):106375. doi: https://doi.org/10.1016/j.cmpb.2021.106375

14. Image Test Time Augmentation with PyTorch [Internet]. TTAch [cited 01.03.2023]. Available from: https://github.com/qubvel/ttach

15. Documentation of the imgaug library for image augmentation [Internet] [cited 01.03.2023]. Available from: https://imgaug.readthedocs.io/en/latest/, free

16. Hussain Z, Gimenez F, Yi D, Rubin D. Differential data augmentation techniques for medical imaging classification tasks. AMIA . Annu Symp proceedings AMIA Symp. 2017. P. 979-984.

17. Chen Y, Yang X-H, Wei Z, et al. Generative Adversarial Networks in Medical Image augmentation: A review. Comput Biol Med. 2022;144:105382. doi: https://doi.org/10.1016/j.compbiomed.2022.105382

18. Shi G, Wang J, Qiang Y, et al. Knowledge-guided synthetic medical image adversarial augmentation for ultrasonography thyroid nodule classification. Comput Methods Programs Biomed. 2020;(196):105611. doi: https://doi.org/10.1016/j.cmpb.2020.105611

19. Neurohive [Internet]. ResNet (34, 50, 101): «остаточные» CNN для классификации изображений. Доступно по: https://neurohive.io/ru/vidy-nejrosetej/resnet-34-50-101/. Ссылка активна на 01.03.2023.

20. ResNet [Internet]. PyTorch Documentation [cited 01.03.2023]. Available from: https://pytorch.org/vision/stable/models/resnet.html

21. Tan M, Le Q V. EfficientNet: Rethinking model scaling for convolutional neural networks. In: 36th International Conference on Machine Learning, ICML 2019; 2019.

22. EfficientNet [Internet]. PyTorch Documentation [cited 01.03.2023]. Available from: https://pytorch.org/vision/stable/models/efficientnet.html

23. Ludwig M, Ludwig B, Mikuła A, et al. The use of artificial intelligence in the diagnosis and classification of thyroid nodules: An update. Cancers (Basel). 2023;15(3):708. doi: https://doi.org/10.3390/cancers15030708