A.A. Egorchev1, D.E. Chikrin2

1, 2 Kazan (Volga Region) Federal University (Kazan, Russia)
1 anton@egorchev.ru, 2 dmltry.kfu@ya.ru

Problem statement. Ultrasound imaging, widely used as a measurement channel in biomedical information-measurement systems, is intrinsically susceptible to acoustic artefacts that distort the visualised information and reduce the reliability of subsequent interpretation. Among the most commonly observed is the mirror-image (doubling) artefact, produced when the ultrasound wave undergoes specular reflection from strongly reflective interfaces and generates a false ghost image on the opposite side of the reflecting boun­dary. Conventional artefact-suppression approaches rely either on manual delineation by an operator, which is time-consuming and poorly reproducible, or on end-to-end neural regression that hides the intermediate localisation stage, impairs interpretability of the processing chain, and complicates integration into existing data-processing pipelines of information-measurement systems.

Goal. The work aims to develop algorithmic support for an information-measurement system of ultrasound visualisation that enables automated detection, localisation and correction of the mirror-image artefact by means of neural network methods while preserving a modular and interpretable processing pipeline.

Results. A two-stage algorithmic pipeline is proposed. The detection and localisation stage combines cascaded preliminary filtering (colour inversion, blurring, histogram equalisation by means of CLAHE, erosion and threshold contour extraction) with geometric analysis based on minimum enclosing circles and pairing of non-overlapping candidate regions. Candidate pairs are then evaluated by a convolutional neural network based on the MobileNet architecture that generates a compact feature representation; the similarity between paired crops is computed via cosine distance over the feature space, and pairs whose score falls within a predefined confidence band are accepted as artefact candidates. The correction stage determines the source region within each accepted pair by brightness analysis, generates a binary inpainting mask and reconstructs the missing content with a generative adversarial network based on the AOT-GAN architecture. The generator comprises an encoder, a sequence of aggregated contextual transformation (AOT) blocks that capture spatial context at multiple dilation rates, and a decoder; the discriminator follows the PatchGAN design. Experimental evaluation on real ultrasound images demonstrates that the proposed pipeline reliably removes the mirror-image artefact while preserving anatomical detail in the surrounding tissue.

Practical relevance. The developed algorithmic support improves the quality and informativeness of ultrasound visualisation data, can be incorporated into the data-processing pipeline of existing information-measurement systems, and lays the foundation for exten­ding the approach to the suppression of other ultrasound artefacts, including comet-tail and ring-down patterns.

Egorchev A.A., Chikrin D.E. Algorithmic support for removal of the mirror-image artefact in information-measurement systems for ultrasound visualisation // Achievements of modern radioelectronics. 2026. V. 80. № 7. P. 30–39. DOI: https://doi.org/10.18127/
j20700784-202607-03

1. Feldman M.K., Katyal S., Blackwood M.S. US artifacts. Radiographics. 2009. Т. 29. № 4. С. 1179–1189.
2. Nasir A.I. Artifact in the Image of Ultrasound. Australian Journal of Basic and Applied Sciences. 2018. V. 12. № 12. P. 131–143.
3. Abu-Zidan F.M., Hefny A.F., Corr P. Clinical ultrasound physics. Journal of Emergencies, Trauma, and Shock. 2011. V. 4. № 4. P. 501–503.
4. Hliboký M., Magyar J., Bundzel M., Petrovčin M., Babič F. Artifact Detection in Lung Ultrasound: An Analytical Approach. Electronics. 2023. V. 12. № 7. P. 1551.
5. Hindi A., Peterson C., Barr R.G. Artifacts in diagnostic ultrasound. Reports in Medical Imaging. 2013. № 6. P. 29–48.
6. Wu W.T., Chang K.V., Mezian K., Naňka O., Yang Y.C., Hsu Y.C., Hsu P.C., Özçakar L. Artifacts in musculoskeletal ultrasonography: from physics to clinics. Diagnostics. 2020. V. 10. № 9. P. 645.
7. Quien M.M., Saric M. Ultrasound imaging artifacts: How to recognize them and how to avoid them. Echocardiography. 2018. V. 35. № 9. P. 1388–1401.
8. Baad M., Lu Z.F., Reiser I., Paushter D. Clinical significance of US artifacts. Radiographics. 2017. V. 37. № 5. P. 1408–1423.
9. Tchelepi H., Ralls P.W. Color comet-tail artifact: clinical applications. American Journal of Roentgenology. 2009. V. 192. № 1. P. 11–18.
10. Oh S.H., Han H.Y., Kim H.J. Comet tail artifact on ultrasonography: is it a reliable finding of benign gallbladder diseases?  Ultrasonography. 2019. V. 38. № 3. P. 221.
11. Lee F.C.Y., Jenssen C., Dietrich C.F. A common misunderstanding in lung ultrasound: the comet tail artefact. Medical Ultrasonography. 2018. V. 20. № 3. P. 379–384.
12. Wang X., Liu D., Song J., Chen X., Yu Y. Quantifying lung ultrasound comets with a convolutional neural network: Initial clinical results. Computers in Biology and Medicine. 2019. V. 107. P. 39–46.
13. Avruch L., Cooperberg P.L. The ring-down artifact. Journal of Ultrasound in Medicine. 1985. V. 4. № 1. P. 21–28.
14. Tsai C.L., Chen W.J., Chang W.T., Wang H.P., Chen S.C., Hsu C.Y. Ring-down artefacts on abdominal sonography to predict pulmonary abnormalities in the emergency department. Emergency Medicine Journal. 2005. V. 22. № 10. P. 747–748.
15. Lucassen R.T., Jafari M.H., Moore N.M., Nelson B.P., Goldsmith A.J., Duhaime E.P., van Sloun R.J.G. Deep learning for detection and localization of B-lines in lung ultrasound. IEEE Journal of Biomedical and Health Informatics. 2023. V. 27. № 9. P. 4352–4361.
16. Cantero-Chinchilla S., Wilcox P.D., Croxford A.J. A deep learning based methodology for artefact identification and suppression with application to ultrasonic images. NDT & E International. 2022. V. 126. P. 102575.
17. Liu S., Wang Y., Yang X., Lei B., Liu L., Li S.X., Ni D., Wang T. Deep learning in medical ultrasound analysis: a review. Engineering. 2019. V. 5. № 2. P. 261–275.
18. Howell L., Ingram N., Lapham R., Morrell A., McLaughlan J.R. Deep learning for real-time multi-class segmentation of artefacts in lung ultrasound. Ultrasonics. 2024. V. 140. P. 107251.
19. Huh J., Khan S., Ye J.C. Ot-driven multi-domain unsupervised ultrasound image artifact removal using a single CNN. arXiv preprint arXiv:2007.05205. 2020.
20. Howard A.G., Zhu M., Chen B., Kalenichenko D., Wang W., Weyand T., Andreetto M., Adam H. MobileNets: Efficient Convolutional Neural Networks for Mobile Vision Applications. arXiv preprint arXiv:1704.04861. 2017.
21. Zeng Y., Fu J., Chao H., Guo B. Aggregated contextual transformations for high-resolution image inpainting. IEEE Transactions on Visualization and Computer Graphics. 2022. V. 29. № 7. P. 3266–3280.
22. Nepomnyashhij O.V., Al`-Sagir M.M.I., Xajdukova V.N., Sharova A.V., Kotov S.A. Ispol`zovanie svyortochnoj nejronnoj seti i glubokogo mashinnogo obucheniya dlya analiza e`lektrokardiogramm. Uspexi sovremennoj radioe`lektroniki. 2024. T. 78. № 3. S. 70–77. DOI: 10.18127/j20700784-202403-08 (in Russian).
23. Chikrin D.E., Smol`nikova K.R., Galiullin I.G. Sovmestnaya rabota: standarty` bezopasnosti dlya kollaborativny`x robotov. Uspexi sovremennoj radioe`lektroniki. 2024. T. 78. № 11. S. 34–40. DOI: 10.18127/j20700784-202411-04 (in Russian).