Artificial Intelligence-assisted Medical Imaging in Interventional Management of Valvular Heart Disease

doi:10.37015/AUDT.2023.230030

Abstract

Abstract:

The integration of medical imaging and artificial intelligence (AI) has revolutionized interventional therapy of valvular heart diseases (VHD), owing to rapid development in multimodality imaging and healthcare big data. Medical imaging techniques, such as echocardiography, cardiovascular magnetic resonance (CMR) and computed tomography (CT), play an irreplaceable role in the whole process of pre-, intra- and post-procedural intervention of VHD. Different imaging techniques have unique advantages in different stages of interventional therapy. Therefore, single imaging technique can’t fully meet the requirements of complicated clinical scenarios. More importantly, a single intraoperative image provides only limited vision of the surgical field, which could be a potential source for unsatisfactory prognosis. Besides, the non-negligible inter- and intra-observer variability limits the precise quantification of heart valve structure and function in daily clinical practice. With the help of analysis clustered and regressed by big data and exponential growth in computing power, AI broken grounds in the interventional therapy of VHD, including preoperative planning, intraoperative navigation, and postoperative follow-up. This article reviews the state-of-the-art progress and directions in the application of AI for medical imaging in the interventional therapy of VHD.

Key words: VHD; AI; Machine learning; Medical imaging

Wenwen Chen, BS, Yuji Xie, MD, Zisang Zhang, MD, Ye Zhu, MS, Yiwei Zhang, MD, Shuangshuang Zhu, MD, PhD, Chun Wu, MD, PhD, Ziming Zhang, MD, Xin Yang, PhD, Man wei Liu, MD, PhD, Mingxing Xie, MD, PhD, Li Zhang, MD, PhD. Artificial Intelligence-assisted Medical Imaging in Interventional Management of Valvular Heart Disease. Advanced Ultrasound in Diagnosis and Therapy, 2023, 7(3): 217-227.

Figures/Tables 4

Figure 1

Table 1

Table 2

AI-assisted medical imaging in pre-operative planning of VHD interventional therapy"

Study	Operation	Task	Technology	Samples	Performance metrics
Theriault-Lauzier, P. et al. [10] (2021)	TAVR	Plane positioning	CNN	94 subjects	Localization error (mm): 0.9 ± 0.8 (testing)
Al, W.A. et al. [11] (2018)	TAVR	Located important anatomic markers	ML	71 subjects	Localization error (mm): 2.04 ± 1.11
Rocatello, G. et al. [12] (2019)	TAVR	Determined the optimal valve size and implantation position	FE	62 subjects	Accuracy (%): 71 Maximum contact pressure (%): 75 Contact pressure index (%): 71
De Jaegere, P. et al. [13] (2016)	TAVR	Simulated the TAVR surgical process	FE	60 subjects	Accuracy (%): 80 Cutoff value (ml/s): 16.0 Sensitivity: 0.72 Specificity: 0.78
Auricchio, F. et al. [14] (2014)	TAVR	Simulated valve implantation	FE	2 subjects	Stress state has consistency (between 2 subjects)
Astudillo, P. et al. [15] (2019)	TAVR	Calculated the size of the implanted prosthesis	CNN	Training: 355subjects Testing: 118 subjects	Total analysis time(s): < 1 Device size has consistency (between the manual and automatic selection)
Astudillo, P. et al. [16] (2019)	TMVR	Measured multiple biological parameters	DL	71 subjects	Total analysis time (s): < 1
Oguz, D. et al. [17] (2019)	TMVR	Explored the correlation between 3D-TEE parameters and MR reduction	3D TEE; Mitral Valve Navigator.	59 subjects	Optimal MR reduction: 68%
Guerrero, M. et al. [18] (2018)	TMVR	Simulated valve implantation	FE	/	Total analysis time: < 3 h
Wang, D.D. et al. [19,20] (2018)	TMVR	Simulated valve implantation	CAD	38 subjects	R2: 0.8169 (neo-LVOT surface area) Sensitivity: 100% Specificity: 96.8%
Kong, F. et al. [21] (2020)	TEER	simulated the biomechanics of MC implantation	FE; MC.	1 subject	Antero-posterior distance: ↓26% Annulus area: ↓19% Valve opening orifice area: ↓48% Regurgitant orifice area: ↓63% Anterior leaflet peak stresses: ↑ 64% Posterior leaflet peak stresses: ↑62% Anterior leaflet peak strains: ↑ 20% Posterior leaflet peak strains: ↑10%
Sturla, F. et al. [22] (2015)	TEER	Simulated the biomechanics of MC implantation	FE; MC.	3 subjects	Systolic CoA: ↑11-40% Systolic leaflet stresses (Kpa): 100-500 Diastolic leaflet stresses (Kpa): 250 (subject 3) Diastolic orifice area (%): ↓58.9%
Caballero, A. et al. [23] (2020)	TEER	Evaluated the biomechanics of MC implantation	FE; FSI; MC.	1 subject	Antero-posterior distance: ↓28% Mitral annulus spherecity index: ↓39% Anatomic regurgitant orifice area: ↓52% Anatomic opening orifice area: ↓71% Diastolic anterior leaflet stress: ↑210% Diastolic posterior leaflet stress: ↑145%
Mansi, T. et al. [24] (2012)	TEER	Evaluated the biomechanical impact of mitral valve repair	FE; ML.	25 subjects	Mean error (mm):1.49 ± 0.62 (ground truth) Mean error (mm):2.75 ± 0.86 (automatic detection) Total analysis time (min): < 14
Dabiri, Y. et al. [25] (2023)	TEER	Predicted the effect of TEER therapy with MC	DL; XGBoost.	1267 FE models	MAPE: 54 and 0.310 (DL) MAPE: 0.115 and 0.231 (XGBoost) Total analysis time (s): < 1

Table 2

Table 3

AI-assisted medical imaging in intra-operative navigation of VHD interventional therapy"

Study	Operation	Task	Technology	Samples	Performance metrics
Biaggi, P. et al. [26] (2020)	TAVR	Investigated the efficacy of FS in the perioperative period of TAVR	FI; EN; 3D TEE.	Total: 138subjects FS+: 69subjects FS-: 69subjects	Procedure time (min): 42.1 ± 15.2 (FS+) Procedure time (min): 49.2 ± 20.7 (FS-) Contrast agent use (ml): 34.3 ± 22.0 (FS+) Contrast agent use (ml): 39.0 ± 23.3 (FS-) Fluoroscopy time (min): 11.4 ± 4.7 (FS+) Fluoroscopy time (min): 10.9 ± 5.5 (FS-) Pearson correlation r: 0.63-0.78 Interclass correlation coefficient: 0.95-0.99
Luo, Z. et al. [27] (2013)	TAVR	Reconstructed aortic valve models and determined the target location for aortic valve prosthesis implantation	MTS; 2D US; 4D CT.	ECG signal	Aortic root segmentation algorithm error (mm): 0.92 ± 0.85 Computational time (ms): 36.13 ± 6.26 Yielding fiducial localization errors (mm): 3.02 ± 0.39 Target registration errors(mm): 3.31 ± 1.55 Deployment distance(mm): 3.23 ± 0.94 Tilting errors (°): 5.85 ± 3.06
Mazomenos, E. B. et al. [28] (2016)	TAVR	Evaluated surgical skills and verified the role of robot assisted TAVR surgery	FE; k-means clustering; EM.	12 subjects (novice group: 6 subjects)	The median value of the procedure time (s): 34.9 (stage 1) The median value of the procedure time (s): 111.2 (stage 2) Maximum accuracy (%): 83 (k-means) Maximum accuracy (%): 91 (EM) Average speed (px/s): 22.3 (stage1) Average speed (px/s): 22 (stage2) P = 0.031(conventional equipment vs robotic system )
Prihadi, E. A. et al. [29] (2018)	TAVR	Quantified aortic ring and root size	3D TEE; AVN.	150 subjects	Mean analysis time (min): 4.2 ± 1.0 r≥0.90 (inter- and intra-observer variability)
Lang, P. et al. [30] (2012)	TAVR	Build TAVI's enhanced image guidance system	3D TEE	/	Mean contour boundary distance error (mm): 1.3 (short-axis views) Mean contour boundary distance error (mm): 2.8 (long-axis views) Mean target registration error (mm): 5.9
Coisne, A. et al. [31] (2020)	TMVR	Defined the optimal 3D TEE parameters for TMVR	3D TEE	57 subjects	AUC: 0.88-0.91 (mitral annular area) AUC: 0.85-0.91 (mitral annular perimeter)
Jin, C. N. et al. [32] (2016)	TMVR	Located MVP	AIUS	90 subjects	Accuracy (%): 89 (nonexperts) Image analysis time (min): 1.9 ± 0.7 (experts) Image analysis time (min): 5.0 ± 0.5 (nonexperts)
Altiok, E. et al. [33] (2011)	TEER	Evaluated the value of RT 3D TEE	RT 3D TEE; 2D TEE.	28 subjects	Advantages: 9/11 (RT 3D TEE)
Melillo, F. et al. [34] (2021)	TEER	Explored the TEER treatment effect after MC implantation	FI; MC.	80 subjects	Fluoroscopy time (min): 37.3 ± 14.6 Procedural time (min): 92.2 ± 36.1
Sündermann, S.H. et al. [35] (2014)	TEER	Evaluated the feasibility and safety of using MC	EN Software; MC.	21 subjects	Radiation dose (Gy/cm2): 146.5 ± 123.6 Total procedure time (min): 136.2 ± 50.2

Table 3

References 44

[1]	Nkomo VT, Gardin JM, Skelton TN, Gottdiener JS, Scott CG, Enriquez-Sarano M. Burden of valvular heart diseases: a population-based study. Lancet 2006; 368:1005-1011. doi: 10.1016/S0140-6736(06)69208-8 pmid: 16980116
[2]	Ouyang D, He B, Ghorbani A, Yuan N, Ebinger J, Langlotz CP, et al. Video-based AI for beat-to-beat assessment of cardiac function. Nature 2020; 580:252-256. doi: 10.1038/s41586-020-2145-8
[3]	Zhang J, Gajjala S, Agrawal P, Tison GH, Hallock LA, Beussink-Nelson L, et al. Fully automated echocardiogram interpretation in clinical practice. Circulation 2018; 138:1623-1635. doi: 10.1161/CIRCULATIONAHA.118.034338 pmid: 30354459
[4]	Kang NG, Suh YJ, Han K, Kim YJ, Choi BW. Performance of prediction models for diagnosing severe aortic stenosis based on aortic valve calcium on cardiac computed tomography: incorporation of radiomics and machine learning. Korean J Radiol 2021; 22:334-343. doi: 10.3348/kjr.2020.0099 pmid: 33236537
[5]	Sengupta PP, Shrestha S, Kagiyama N, Hamirani Y, Kulkarni H, Yanamala N, et al. A machine-learning framework to identify distinct phenotypes of aortic stenosis severity. JACC Cardiovasc Imaging 2021; 14:1707-1720. doi: 10.1016/j.jcmg.2021.03.020 pmid: 34023273
[6]	Yang F, Chen X, Lin X, Chen X, Wang W, Liu B, et al. Automated analysis of Doppler echocardiographic videos as a screening tool for valvular heart diseases. JACC Cardiovascular Imaging 2022; 15:551-563. doi: 10.1016/j.jcmg.2021.08.015
[7]	Moghaddasi H, Nourian S. Automatic assessment of mitral regurgitation severity based on extensive textural features on 2D echocardiography videos. Comput Biol Med 2016; 73:47-55. doi: 10.1016/j.compbiomed.2016.03.026 pmid: 27082766
[8]	Pimor A, Galli E, Vitel E, Corbineau H, Leclercq C, Bouzille G, et al. Predictors of post-operative cardiovascular events, focused on atrial fibrillation, after valve surgery for primary mitral regurgitation. Eur Heart J Cardiovasc Imaging 2019; 20:177-184. doi: 10.1093/ehjci/jey049 pmid: 29608669
[9]	Bartko PE, Heitzinger G, Spinka G, Pavo N, Prausmüller S, Kastl S, et al. Principal morphomic and functional components of secondary mitral regurgitation. JACC Cardiovascular Imaging 2021; 14:2288-2300. doi: 10.1016/j.jcmg.2021.05.020
[10]	Theriault-Lauzier P, Alsosaimi H, Mousavi N, Buithieu J, Spaziano M, Martucci G, et al. Recursive multiresolution convolutional neural networks for 3D aortic valve annulus planimetry. Int J Comput Assist Radiol Surg 2020; 15:577-588. doi: 10.1007/s11548-020-02131-0 pmid: 32130646
[11]	Al WA, Jung HY, Yun ID, Jang Y, Park HB, Chang HJ. Automatic aortic valve landmark localization in coronary CT angiography using colonial walk. PLoS One 2018; 13:e0200317. doi: 10.1371/journal.pone.0200317
[12]	Rocatello G, El Faquir N, de Backer O, Swaans MJ, Latib A, Vicentini L, et al. The impact of size and position of a mechanical expandable transcatheter aortic valve: novel insights through computational modelling and simulation. J Cardiovasc Transl Res 2019; 12:435-446. doi: 10.1007/s12265-019-09877-2 pmid: 31444672
[13]	de Jaegere P, De Santis G, Rodriguez-Olivares R, Bosmans J, Bruining N, Dezutter T, et al. Patient-specific computer modeling to predict aortic regurgitation after transcatheter aortic valve replacement. JACC Cardiovasc Interv 2016; 9:508-512. doi: 10.1016/j.jcin.2016.01.003
[14]	Auricchio F, Conti M, Morganti S, Reali A. Simulation of transcatheter aortic valve implantation: a patient-specific finite element approach. Comput Methods Biomech Biomed Engin 2014; 17:1347-1357. doi: 10.1080/10255842.2012.746676 pmid: 23402555
[15]	Astudillo P, Mortier P, Bosmans J, De Backer O, de Jaegere P, De Beule M, et al. Enabling automated device size selection for transcatheter aortic valve implantation. J Interv Cardiol 2019; 2019:3591314.
[16]	Astudillo P, De Beule M, Dambre J, Mortier P. Towards safe and efficient preoperative planning of transcatheter mitral valve interventions. Morphologie 2019; 103:139-147. doi: S1286-0115(19)30053-0 pmid: 31570309
[17]	Oguz D, Eleid MF, Dhesi S, Pislaru SV, Mankad SV, Malouf JF, et al. Quantitative three-dimensional echocardiographic correlates of optimal mitral regurgitation reduction during transcatheter mitral valve repair. J Am Soc Echocardiogr 2019; 32:1426-1435. doi: 10.1016/j.echo.2019.06.014
[18]	Guerrero M, Urena M, Wang DD, O'Neill W, Feldman T. Reply: patient-specific computer modeling for the planning of transcatheter mitral valve replacement. J Am Coll Cardiol 2018; 72:958. doi: S0735-1097(18)35280-X pmid: 30115239
[19]	Wang DD, Eng MH, Greenbaum AB, Myers E, Forbes M, Karabon P, et al. Validating a prediction modeling tool for left ventricular outflow tract (LVOT) obstruction after transcatheter mitral valve replacement (TMVR). Catheter Cardiovasc Interv 2018; 92:379-387. doi: 10.1002/ccd.v92.2
[20]	Wang DD, Eng M, Greenbaum A, Myers E, Forbes M, Pantelic M, et al. Predicting LVOT obstruction after TMVR. JACC Cardiovasc Imaging 2016; 9:1349-1352. doi: S1936-878X(16)30041-9 pmid: 27209112
[21]	Kong F, Caballero A, McKay R, Sun W. Finite element analysis of MitraClip procedure on a patient-specific model with functional mitral regurgitation. J Biomech 2020; 104:109730. doi: 10.1016/j.jbiomech.2020.109730
[22]	Sturla F, Redaelli A, Puppini G, Onorati F, Faggian G, Votta E. Functional and biomechanical effects of the edge-to-edge repair in the setting of mitral regurgitation: consolidated knowledge and novel tools to gain insight into its percutaneous implementation. Cardiovasc Eng Technol 2015; 6:117-140. doi: 10.1007/s13239-014-0208-4 pmid: 26577231
[23]	Caballero A, Mao W, McKay R, Hahn RT, Sun W. A comprehensive engineering analysis of left heart dynamics after MitraClip in a functional mitral regurgitation patient. Front Physiol 2020; 11:432. doi: 10.3389/fphys.2020.00432 pmid: 32457650
[24]	Mansi T, Voigt I, Georgescu B, Zheng X, Mengue EA, Hackl M, et al. An integrated framework for finite-element modeling of mitral valve biomechanics from medical images: application to MitralClip intervention planning. Med Image Anal 2012; 16:1330-1346. doi: 10.1016/j.media.2012.05.009 pmid: 22766456
[25]	Dabiri Y, Mahadevan VS, Guccione JM, Kassab GS. Machine learning used for simulation of MitraClip intervention: A proof-of-concept study. Front Genet 2023; 14:1142446. doi: 10.3389/fgene.2023.1142446
[26]	Biaggi P, Sager DF, Külling J, Küest S, Wyss C, Hürlimann D, et al. Potential value of fusion imaging and automated three-dimensional heart segmentation during transcatheter aortic valve replacement. J Am Soc Echocardiogr 2020;33:516-517.
[27]	Luo Z, Cai J, Peters TM, Gu L. Intra-operative 2-D ultrasound and dynamic 3-D aortic model registration for magnetic navigation of transcatheter aortic valve implantation. IEEE Trans Med Imaging 2013; 32:2152-2165. doi: 10.1109/TMI.2013.2275233
[28]	Mazomenos EB, Chang PL, Rippel RA, Rolls A, Hawkes DJ, Bicknell CD, et al. Catheter manipulation analysis for objective performance and technical skills assessment in transcatheter aortic valve implantation. Int J Comput Assist Radiol Surg 2016; 11:1121-1131.
[29]	Prihadi EA, van Rosendael PJ, Vollema EM, Bax JJ, Delgado V, Ajmone Marsan N,. Feasibility, accuracy, and reproducibility of aortic annular and root sizing for transcatheter aortic valve replacement using novel automated three-dimensional echocardiographic software: comparison with multi-detector row computed tomography. J Am Soc Echocardiogr 2018; 31:505-514. doi: 10.1016/j.echo.2017.10.003
[30]	Lang P., Rajchl M., McLeod A. J., Chu M. & Peters T. Feature identification for image-guided transcatheter aortic valve implantation. Medical Imaging 2012.
[31]	Coisne A, Pontana F, Aghezzaf S, Mouton S, Ridon H, Richardson M, et al. Utility of three-dimensional transesophageal echocardiography for mitral annular sizing in transcatheter mitral valve replacement procedures: a cardiac computed tomographic comparative study. J Am Soc Echocardiogr 2020; 33:1245-1252. doi: 10.1016/j.echo.2020.04.030
[32]	Jin CN, Salgo IS, Schneider RJ, Kam KK, Chi WK, So CY, et al. Using anatomic intelligence to localize mitral valve prolapse on three-dimensional echocardiography. J Am Soc Echocardiogr 2016; 29:938-945.
[33]	Altiok E, Becker M, Hamada S, Reith S, Marx N, Hoffmann R. Optimized guidance of percutaneous edge-to edge repair of the mitral valve using real-time 3-D transesophageal echocardiography. Clin Res Cardiol 2011; 100:675-681. doi: 10.1007/s00392-011-0296-1 pmid: 21369924
[34]	Melillo F, Fisicaro A, Stella S, Ancona F, Capogrosso C, Ingallina G, et al. Systematic fluoroscopic-echocardiographic fusion imaging protocol for transcatheter edge-to-edge mitral valve repair intraprocedural monitoring. J Am Soc Echocardiogr 2021; 34:604-613. doi: 10.1016/j.echo.2021.01.010
[35]	Sündermann SH, Biaggi P, Grünenfelder J, Gessat M, Felix C, Bettex D, et al. Safety and feasibility of novel technology fusing echocardiography and fluoroscopy images during MitraClip interventions. EuroIntervention 2014;9:1210-1216.
[36]	Navarese EP, Zhang Z, Kubica J, Andreotti F, Farinaccio A, Bartorelli AL, et al. Development and validation of a practical model to identify patients at risk of bleeding after TAVR. JJACC Cardiovasc Interv 2021; 14:1196-1206.
[37]	Jia Y, Luosang G, Li Y, Wang J, Li P, Xiong T, et al. Deep learning in prediction of late major bleeding after transcatheter aortic valve replacement. Clin Epidemiol 2022; 14:9-20. doi: 10.2147/CLEP.S333147 pmid: 35046728
[38]	Kwak S, Everett RJ, Treibel TA, Yang S, Hwang D, Ko T, et al. Markers of myocardial damage predict mortality in patients with aortic stenosis. J Am Coll Cardiol 2021; 78:545-558. doi: 10.1016/j.jacc.2021.05.047 pmid: 34353531
[39]	Zweck E, Spieker M, Horn P, Iliadis C, Metze C, Kavsur R, et al. Machine learning identifies clinical parameters to predict mortality in patients udergoing transcatheter mitral valve repair. JACC Cardiovasc Interv 2021; 14:2027-2036. doi: 10.1016/j.jcin.2021.06.039
[40]	Modine T, Perrin N, Ben Ali W. Trust in machine learning models for mortality prediction following mitral TEER: Are we ready yet? JACC Cardiovasc Interv 2021; 14:2037-2038. doi: 10.1016/j.jcin.2021.08.002
[41]	Hernandez-Suarez DF, Kim Y, Villablanca P, Gupta T, Wiley J, Nieves-Rodriguez BG, et al. Machine learning prediction models for in-hospital mortality after transcatheter aortic valve replacement. JACC Cardiovasc Interv 2019; 12:1328-1338. doi: 10.1016/j.jcin.2019.06.013
[42]	Tse G, Zhou J, Lee S, Liu Y, Leung KSK, Lai RWC, et al. Multi-parametric system for risk stratification in mitral regurgitation: A multi-task Gaussian prediction approach. Eur J Clin Invest 2020; 50:e13321. doi: 10.1111/eci.v50.11
[43]	Engelhardt S, Sauerzapf S, Brčić A, Karck M, Wolf I, De Simone R. Replicated mitral valve models from real patients offer training opportunities for minimally invasive mitral valve repair. Interact Cardiovasc Thorac Surg 2019; 29:43-50. doi: 10.1093/icvts/ivz008
[44]	Liu J, Al'Aref SJ, Singh G, Caprio A, Moghadam AAA, Jang SJ, et al. An augmented reality system for image guidance of transcatheter procedures for structural heart disease. PloS One 2019; 14:e0219174. doi: 10.1371/journal.pone.0219174

Study	VHD	Task	AI methods	Samples	Performance metrics
Kang, N. G. et al. [4](2021)	AS	Diagnosed severe AS	LASSO; RFs; XGBoost	Training: 312 subjects Testing: 96 subjects	AUC: 0.921 (LASSO & XGBoost)
Sengupta, P. P. et al. [5](2021)	AS	Distinguished AS phenotypes	TDA; ML	1964 subjects	AUC: 0.988 Accuracy (%): 94.3 Precision (%): 91.3 Recall (%): 95.5
Yang, F. et al. [6](2014)	AS	Differential diagnosis with other diseases	ML	Training: 1335 subjects Validation: 311 subjects Testing: 434 subjects	AUC: 0.97 Accuracy (%): 94 Sensitivity (%): 90 Specificity (%): 94
Moghaddasi, H. et al. [7](2016)	MR	Detected normal, mild, moderate and severe MR subjects	Textural analysis; SVM; LDA; TM	5004 images	Accuracy (%): 99.45 (SVM) Accuracy (%): 95.72 (LDA,NN) Accuracy (%): 95 (TM) Sensitivity (%): 99.38 Specificity (%): 99.63
Pimor, A. et al. [8](2019)	MR	Distinguished MR phenotypes	DL	122 subjects	HR: 3.57 (1.72-7.44)
Bartko, P. E. et al. [9](2021)	MR	Explored the morphological and Functional characteristics of secondary MR	PC; cluster analysis	383 subjects	HR: 2.18 (clusters3)