A.G. Levchuk1, V.A. Fokin2, A.V. Ryzhkov3, M.S. Baev4, D. Bendahan5, W. Al-Haidri6, E.A. Brui7

1,6,7 ITMO University (Saint-Petersburg, Russian)
2–4 Almazov National Medical Research Centre (Saint-Petersburg, Russian)
5 Aix-Marseille University, National Center for Scientific Research, Biological and Medical Magnetic Resonance Center (Marseille, France)
1anatolii.levchuk@metalab.ifmo.ru, 2vladfokin@mail.ru, 3abanderos83@mail.ru, 4mikael.baev@mail.ru, 5david.bendahan@univ-amu.fr, 6waleed.al.haidri@metalab.ifmo.ru, 7katya.bruy@gmail.com

The improvement of technologies for the quantitative characterization of myocardial fibrosis has great prospects for improving the prediction of the outcomes of cardiovascular diseases, as well as for the choice of therapeutic and surgical strategies. In clinical practice, for the subsequent quantitative assessment of fibrosis, magnetic resonance (MR) images of the heart are usually processed either manually or semi-automatically (threshold methods).

In this paper, the effect of the method of preliminary preparation of two-dimensional post-contrast MR images of the heart of patients with postinfarction cardiosclerosis on the effectiveness of neural network segmentation of the left ventricular myocardium and fibrous tissue is investigated. In addition, this work is aimed at automating the stage of image preparation, and creating a fully automatic segmentation method.

As a result of the work, a data set was created consisting of MR images of the heart with delayed contrast of patients with signs of postinfarction cardiosclerosis with marked structures of a healthy myocardium and left ventricular cavity, as well as myocardial fibrosis. Several neural network models were trained on the created dataset in order to automate the calculation of the relative volume of left ventricular fibrosis.

As part of the work, it was shown that manual image preparation provides high-precision segmentation of left ventricular fibrosis by a neural network with U-Net architecture. At the same time, several options for such preparation were investigated and the most optimal ones were identified, ensuring the similarity of predicted and reference fibrosis masks at a level above 85%. An attempt to automate the pre-training steps led to a slight decrease in similarity (up to 74%). However, the target metric – the relative volume of fibrosis – in the most effective version of the automatic algorithm showed a high correlation (p = 0,91; p ≤ 0,001) with that obtained manually by an experienced radiologist.

The proposed automatic method, which provides radiologists with masks of fibrosis and healthy myocardium, can be used as a decision support system.

Levchuk A.G., Fokin V.A., Ryzhkov A.V., Baev M.S., Bendahan D., Al-Haidri W., Brui E.A. Automatic and semi-automatic segmentation method of post-myocardial infarction according to magnetic resonance imaging with late gadolinium enhancement. Biomedicine Radioengineering. 2024. V. 27. № 3. P. 13–27. DOI: https:// doi.org/10.18127/j15604136-202403-02 (In Russian)

1. Wu K.C., Weiss R.G., Thiemann D.R., Kitagawa K., Schmidt A., Dalal D., Lai S., Bluemke D.A., Gerstenblith G., Marban E., Tomaselli G.F., Lima J.A. Late gadolinium enhancement by cardiovascular magnetic resonance heralds an adverse prognosis in nonischemic cardiomyopathy.
   J. Am Coll Cardiol. 2008. 51(25):2414–21.
2. Kryukov N.A. Direct revascularization of the myocardium in patients with coronary heart disease complicated by a marked decrease in the contractility of the left ventricular myocardium, Dissertation for the degree of Candidate of Medical Sciences. Almazov National Medical Research Centre. Saint-Petersburg. 2021.
3. Wong T.C., Piehler K., Meier C.G., Testa S.M., Klock A.M., Aneizi A.A., Shakesprere J., Kellman P., Shroff S.G., Schwartzman D.S., Mulukutla S.R., Simon M.A., Schelbert E.B. Association between extracellular matrix expansion quantified by cardiovascular magnetic resonance and short-term mortality. Circulation. 2012. 126(10), 1206–1216. https://doi.org/10.1161/CIRCULATIONAHA.111.089409
4. Shivkumar K., Qu Z., Harvey R. Cardiac fibrosis in three dimensions – mechanistic insights into arrhythmic risk due to hypertrophy. J Physiol. 2023. 601: 249–250. https://doi.org/10.1113/JP283710
5. Kim R.J., Fieno D.S., Parrish T.B., Harris K., Chen E.L., Simonetti O., Bundy J., Finn J.P., Klocke F.J., Judd R.M. Relationship of MRI delayed contrast enhancement to irreversible injury, infarct age, and contractile function. Circulation. 1999. 100(19), 1992–2002. https://doi.org/10.1161/01.cir.100.19.1992
6. Jung J., Kim Y.-H., Kim N., Yang D.H. Patient-specific 17-segment myocardial modeling on a bull's-eye map. Journal of Applied Clinical Medical Physics. 2016. 17: 453-465. https://doi.org/10.1120/jacmp.v17i5.6237
7. Cerqueira M.D., Weissman N.J., Dilsizian V., Jacobs A.K., Kaul S., Laskey W.K., Pennell D.J., Rumberger J.A., Ryan T., Verani M.S. Standardized Myocardial Segmentation and Nomenclature for Tomographic Imaging of the Heart. Journal of Cardiovascular Magnetic Resonance – J CARDIOVASC MAGN RESON. 2002. 4. 203–210. 10.1081/JCMR-120003946.
8. Sinicyn V.E., Stukalova O.V., Larina O.V., Smirnova M.D., Ageev F.T., Ternovoj S.K. Magnitno-rezonansnaya tomografiya serdcza s otsrochenny`m kontrastirovaniem v ocenke ochagovy`x izmenenij v miokarde levogo zheludochka u bol`ny`x s gipertroficheskoj kardiomiopatiej. Kardiologicheskij vestnik. 2008. T. 3, № 2(15). S. 54–57. EDN QYVJIL.
9. Simonetti O.P., Kim R.J., Fieno D.S., Hillenbrand H.B., Wu E., Bundy J.M., Finn J.P., Judd R.M. An improved MR imaging technique for the visualization of myocardial infarction. Radiology. 2001 Jan. 218(1): 215-23. doi: 10.1148/radiology.218.1.r01ja50215. PMID: 11152805.
10. Flett A.S., Hasleton J., Cook C., Hausenloy D., Quarta G., Ariti C., Muthurangu V., Moon J.C. Evaluation of techniques for the quantification of myocardial scar of differing etiology using cardiac magnetic resonance. JACC Cardiovasc Imaging. 2011 Feb. 4(2):150-6. doi: 10.1016/j.jcmg.2010.11.015. PMID: 21329899.
11. Stukalova O.V., Aparina O.P., Mironova N.A., Golicyn S.P. Fibroz miokarda levogo predserdiya po danny`m magnitno-rezonansnoj tomografii s otsrochenny`m kontrastirovaniem u bol`ny`x fibrillyaciej predserdij. Al`manax klinicheskoj mediciny`. 2015. № 43. S. 29-37. EDN VCNWXD.
12. Carminati M.C., Boniotti C., Fusini L., Andreini D., Pontone G., Pepi M., Caiani E.G. Comparison of Image Processing Techniques for Nonviable Tissue Quantification in Late Gadolinium Enhancement Cardiac Magnetic Resonance Images. J Thorac Imaging. 2016 May. 31(3):168-76. doi: 10.1097/RTI.0000000000000206. PMID: 27043426.
13. Muscogiuri G., Rehwald W.G., Schoepf U.J., Suranyi P., Litwin S.E., De Cecco C.N., Wichmann J.L., Mangold S., Caruso D., Fuller S.R., Bayer Nd. R.R., Varga-Szemes A. T(Rho) and magnetization transfer and INvErsion recovery (TRAMINER)-prepared imaging: A novel contrast-enhanced flow-independent dark-blood technique for the evaluation of myocardial late gadolinium enhancement in patients with myocardial infarction. J Magn Reson Imaging. 2017 May. 45(5):1429-1437. doi: 10.1002/jmri.25498. Epub 2016 Sep 30. PMID: 27690324.
14. Berdibekov B.Sh., Aleksandrova S.A., Goluxova E.Z. Kolichestvennaya ocenka miokardial`nogo fibroza s primeneniem magnitno-rezo­nansnoj tomografii s otsrochenny`m kontrastirovaniem pri nekoronarogenny`x zheludochkovy`x aritmiyax. Kreativnaya kardiologiya. 2021. 15 (3): 342–53. DOI: 10.24022/1997-3187-2021-15-3-342-353
15. Fel J.T., Ellis C.T., Turk-Browne N.B. Automated and manual segmentation of the hippocampus in human infants. Dev Cogn Neurosci. 2023 Apr. 60:101203. doi: 10.1016/j.dcn.2023.101203. Epub 2023 Jan 27. PMID: 36791555; PMCID: PMC9957787.
16. Zabihollahy F., White J.A., Ukwatta E. Myocardial scar segmentation from magnetic resonance images using convolutional neural network. Proceedings V. 10575, Medical Imaging 2018: Computer-aided diagnosis.
17. Zabihollahy F., White J.A., Ukwatta E. Convolutional neural network-based approach for segmentation of left ventricle myocardial scar from 3D late gadolinium enhancement MR images. Med Phys. 2019. 46(4):1740–51. https://doi.org/10.1002/mp.13436.
18. Dice L.R. Measures of the Amount of Ecologic Association Between Species. Ecology.1945. 26 (3): 297–302.
19. Sørensen T. A method of establishing groups of equal amplitude in plant sociology based on similarity of species and its application to analyses of the vegetation on Danish commons. Kongelige Danske Videnskabernes Selskab. 1948. 5 (4): 1–34.
20. Zabihollahy F., Rajchl M., White A.J., Ukwatta E. Fully automated segmentation of left ventricular scar from 3D late gadolinium enhancement magnetic resonance imaging using a cascaded multi-planar U-Net (CMPU-Net). Medical Physics. 2020.
21. Lustermans D.R.P.R.M., Amirrajab S., Veta M., Breeuwer M., Scannell C.M. Optimized automated cardiac MR scar quantification with GAN-based data augmentation. Comput Methods Programs Biomed. 2022 Nov. 226: 107116. doi: 10.1016/j.cmpb.2022.107116. Epub 2022 Sep 7. PMID: 36148718.
22. MedSeg, https://www.medseg.ai/ [Accessed: July. 17, 2023].
23. Lalande A., Chen Z., Decourselle T., Qayyum A., Pommier T., Lorgis L., de la Rosa E., Cochet A., Cottin Y., Ginhac D., Salomon M., Couturier R., Meriaudeau F. Emidec: A Database Usable for the Automatic Evaluation of Myocardial Infarction from Delayed-Enhancement Cardiac MRI. Data 2020, 5, 89. doi: https://doi.org/10.3390/data5040089
24. Ronneberger O., Fischer Ph., Brox Th. U-Net: Convolutional Networks for Biomedical Image Segmentation. 2015.
25. Oktay O., Schlemper J., Folgoc L.L., Lee M.J., Heinrich M.P., Misawa K., Mori K., McDonagh S.G., Hammerla N.Y., Kainz B., Glocker B., Rueckert D. Attention U-Net: Learning Where to Look for the Pancreas. ArXiv. 2018. abs/1804.03999.
26. Tarekegn A.N., Giacobini M., Michalak K. A review of methods for imbalanced multi-label classification, Pattern Recognition. 2021. V. 118.
27. van Lieshout F.E., Klein R.C., Kolk M.Z., van Geijtenbeek K., Vos R., Ruiperez-Campillo S., Feng R., Deb B., Ganesan P., Knops R., Isgum I., Narayan S., Bekkers E., Vos B., Tjong F.V.Y.  Deep Learning for Ventricular Arrhythmia Prediction Using Fibrosis Segmentations on Cardiac MRI Data. In 2022 Computing in Cardiology. 2022. CinC 2022 (Vol. 2022-September). (Computing in Cardiology). IEEE Computer Society. https://doi.org/10.22489/CinC.2022.191
28. Segment CMR, https://medviso.com/cmr/ [Accessed: January. 23, 2023].
29. Cain P.A., Ugander M., Palmer J., Carlsson M., Heiberg E., Arheden H. Quantitative polar representation of left ventricular myocardial perfusion, function and viability using SPECT and cardiac magnetic resonance: initial results. Clinical Physiology and Functional Imaging. 2005. 25: 215-222. https://doi.org/10.1111/j.1475-097X.2005.00618.x
30. CMRSegTools, https://github.com/OpenCardiacMRISoftware/CMRSegTools [Accessed: January. 23, 2023].
31. Romero R.W.A., Viallon M., Spaltenstein J., Petrusca L., Bernard O., Belle L., Clarysse P., Croisille P. CMRSegTools: An open-source software enabling reproducible research in segmentation of acute myocardial infarct in CMR images. PloS one. 2022. 17(9), e0274491. https://doi.org/10.1371/journal.pone.0274491
32. Engblom H., Tufvesson J., Jablonowski R., Carlsson M., Aletras A.H., Hoffmann P., Jacquier A., Kober F., Metzler B., Erlinge D., Atar D., Arheden H., Heiberg E. A new automatic algorithm for quantification of myocardial infarction imaged by late gadolinium enhancement cardiovascular magnetic resonance: experimental validation and comparison to expert delineations in multi-center, multi-vendor patient data. Journal of cardiovascular magnetic resonance: official journal of the Society for Cardiovascular Magnetic Resonance. 2016. 18(1), 27. https://doi.org/10.1186/s12968-016-0242-5
33. EMIDEC, https://emidec.com [Accessed: February. 20, 2024].