Artificial Intelligence in Identifying Disease States: History
View Latest Version
Please note this is an old version of this entry, which may differ significantly from the current revision.
Contributor:
Timothy Barry
,
Juan Maria Farina
,
Chieh-Ju Chao
,
Chadi Ayoub
,
Jiwoong Jeong
,
Bhavik N. Patel
,
Imon Banerjee
,
Reza Arsanjani

Echocardiography is an integral part of the diagnosis and management of cardiovascular disease. The use and application of artificial intelligence (AI) is a rapidly expanding field in medicine to improve consistency and reduce interobserver variability. AI can be successfully applied to echocardiography in addressing variance during image acquisition and interpretation. Furthermore, AI and machine learning can aid in the diagnosis and management of cardiovascular disease. In the realm of echocardiography, accurate interpretation is largely dependent on the subjective knowledge of the operator. Echocardiography is burdened by the high dependence on the level of experience of the operator, to a greater extent than other imaging modalities like computed tomography, nuclear imaging, and magnetic resonance imaging. AI technologies offer new opportunities for echocardiography to produce accurate, automated, and more consistent interpretations. 

  • echocardiography
  • machine learning
  • artificial intelligence

1. Introduction

The rationale behind the use of AI in echocardiography to identify disease states is based on its capacity to automatically analyze features from images and data that are beyond human perception [1]. During routine echocardiography, a huge volume of potentially diagnostic information could be underutilized, considering that the totality of data generated can be hard to interpret by human experts in a short time period [2]. AI can help identify the true value of these undiscovered findings and can analyze this information faster than human experts. Therefore, the potential clinical applications of AI in echocardiography are rapidly increasing, including the identification of specific disease states and processes, such as valvular heart diseases, coronary artery disease, hypertrophic cardiomyopathy, cardiac amyloidosis, cardiomyopathies, and cardiac masses.

2. Valvular Heart Disease

In the field of valvular heart diseases, the focus of AI has been on the echocardiographic quantification of the severity of valve disorders and the identification of high-risk populations [1]. Using image recognition algorithms, valve disease states have been directly detected from raw images, but images have also been integrated with clinical information to identify new predictors of disease progression. Previous studies developed highly accurate algorithms based on images that could establish the severity of mitral and aortic valve disease, recognize the presence of prosthetic valves, and identify rheumatic heart disease [3][4][5]. Further progression in this field could transform how patients with valve diseases are evaluated and managed, as deep learning algorithms could simulate or replace the multimodal evaluation currently required [1].
In a recent study including almost 2000 patients with aortic stenosis, AI integrated echocardiography measurements to improve the classification of disease severity and to identify high-risk subgroups [6]. The identification of higher-risk subjects (higher aortic valve calcium scores, larger late gadolinium enhancement, higher biomarker levels, and greater incidences of negative clinical outcomes) has the potential to optimize the timing of aortic valve replacements [6]. In another recent publication including a large training (n = 1335) and validated (n = 311) cohort, a framework for the automatic screening of echocardiographic videos for mitral and aortic disease was developed [7]. This deep learning algorithm was able to classify echocardiographic views, detect the presence of valve heart disease, and quantify disease severity with high accuracy (AOC > 0.88 for all left heart valve diseases) [7]. These novel findings support the effectiveness of an automated framework, trained on routine echocardiographic datasets, to screen, classify, and quantify the severity of conditions that are frequent in medical practice.

3. Coronary Artery Disease

Cardiac imaging is key for the effective management of patients with coronary artery disease [8]. However, regional wall motion abnormalities traditionally need to be subjectively identified by operators, and interobserver and intraobserver variability can be high [9]. To overcome this issue, an automated image processing pipeline was recently developed to extract geometric and kinematic features from stress echocardiograms [10]. This machine learning model obtained high classification accuracy (specificity of 92.7% and a sensitivity of 84.4%) for the identification of patients with severe coronary artery disease [10]. These results support the use of AI for the analysis of stress echocardiograms to provide automated classifications and to improve accuracy, inter-reader agreement, and reader confidence. Moreover, these findings are especially important when considering that the interpretation of stress echocardiography is widely recognized as one of the most challenging activities for echocardiographers [11].
Another potential implementation of AI in the field of coronary artery disease could be the differentiation between diseases that commonly present with signs and symptoms similar to an acute coronary syndrome. In that sense, a novel cohort study developed a real-time system for fully automated interpretation of echocardiogram videos to differentiate TakoTsubo syndrome from acute myocardial infarction [12]. While this model demonstrated to be more accurate than expert cardiologists in echocardiography-based disease classification, further studies are needed before clinical application.
Lastly, AI models could potentially provide a prediction of left ventricular recovery after coronary syndromes. One study developed a method based on the texture parameters of echocardiograms to evaluate left ventricular function recovery one year after myocardial infarction [13]. Even though the preliminary results were promising (the estimated prediction error was lower than 30%), further studies are warranted for clinical application.

4. Etiology Determination of Increased Left Ventricular Wall Thickness

In cases of increased left ventricular wall thickness, conventional echocardiography may be not sufficient for the etiological diagnoses, and more complex imaging modalities are usually needed. Myocardial texture is generally difficult to assess and quantify in routine echocardiography using only the human impression [14]. One study used echocardiography-AI-based myocardial texture analysis to differentiate hypertrophic cardiomyopathy, hypertensive heart disease, and uremic cardiomyopathy [15]. Hypertrophic cardiomyopathy showed the most homogeneous myocardial texture and was significantly different from the other diagnosis, thus supporting AI-based myocardial texture features as a potential approach to left ventricle hypertrophy etiology differentiation.
Another study investigated the diagnostic value of a machine learning framework that incorporates echocardiographic data for automated discrimination of hypertrophic cardiomyopathy from physiological hypertrophy seen in athletes [16]. This AI model showed increased sensitivity and specificity compared with conventional parameters, suggesting that the use of echocardiography images in machine learning algorithms can assist in the discrimination of physiological versus pathological patterns of hypertrophic remodeling.
Cardiac amyloidosis is characterized by left ventricular hypertrophy and can mimic hypertrophic cardiomyopathy. The impact of cardiac amyloidosis on cardiovascular imaging has been widely described, but isolated echocardiography findings have not been sufficiently specific or sensitive to be used as definitive diagnostic tools for this disease. Recently, a video-based echocardiography model for cardiac amyloidosis using only the apical four-chamber view demonstrated very good performance (C-statistics of 0.96) and outperformed expert human readers in a study including five academic medical centers across two countries [17]. Overall, the model’s superior performance was more apparent for Transthyretin Amyloidosis (ATTR) than AL amyloidosis. A second cohort study developed an AI-guided workflow that automatically quantified left ventricle wall thickness on echocardiography while also predicting the cause of left ventricle hypertrophy as either hypertrophic cardiomyopathy or cardiac amyloidosis [18]. This deep learning model accurately identified subtle changes in left ventricle wall geometric measurements and the causes of hypertrophy, thus providing a more efficient clinical evaluation of this group of patients.
In one additional study, authors developed a deep learning algorithm for the differential diagnosis of common left ventricular hypertrophy etiologies (hypertensive heart disease, hypertrophic cardiomyopathy, and AL-cardiac amyloidosis). A convolutional neural network long short-term memory algorithm was constructed to classify the three diagnoses using five standard echo views (parasternal long-axis, parasternal short-axis, apical four-chamber, apical two-chamber, and apical three-chamber). The study population included a training (n = 620), a validation (n = 155), and a test cohort (n = 155). In the test cohort, the Area under the curve (AUC) for the AI model was 0.962 for hypertensive heart disease, 0.982 for hypertrophic cardiomyopathy, and 0.996 for AL-cardiac amyloidosis. The overall diagnostic accuracy was significantly higher for the deep learning algorithm than for echocardiography specialists, therefore supporting that the use of AI can improve the diagnostic process in patients with left ventricular hypertrophy [19].

5. Cardiomyopathies

AI-assisted diagnosis of cardiomyopathies can be based on ventricular segmentation, measurement of volumes, and automatic assessment of myocardial function and motion [13][14]. One of the most significant benefits of AI in this field may be the potential improved diagnostic performance, particularly in the early stages of some cardiomyopathies where no obvious structural echocardiographic signs may be detected by human perception [14].
Automatic detection of dilated cardiomyopathy from echocardiography videos has been proposed by previous studies. A machine learning framework based on support vector machines was used in one study to separate normal from dilated left ventricles [20]. Even though the performance of the classification showed promising results (classification accuracy was 78%), more information is needed before considering clinical application [20].
Deep learning algorithms were developed to distinguish specific cardiomyopathies using echocardiography movies. One study used AI-assisted diagnosis to differentiate cardiac sarcoidosis from healthy subjects. The diagnostic accuracy of this AI algorithm based on echocardiography videos was not significantly different from the interpretation of the echocardiography movies by human experts [21]. A more recent study proposed a machine learning algorithm based on clinical and speckle-tracking echocardiography data to distinguish between constrictive pericarditis and restrictive cardiomyopathy [22].
A recent study reported an end-to-end deep learning framework that differentiates four common cardiovascular diseases (Atrial Septal Defect, Dilated Cardiomyopathy, Hypertrophic Cardiomyopathy, and prior Myocardial Infarction) from normal subjects. Interestingly, this included 1807 echocardiographic videos obtained during standard clinical care of patients from ultrasound equipment from several different manufacturers and models, thus broadening the application of AI-assisted echocardiography in different medical settings. Moreover, the algorithm identified anatomic regions of interest relevant to each diagnosis, in a similar fashion to an echocardiographer’s approach to interpretation (interatrial septum for atrial septal defect, the left ventricular chamber for dilated cardiomyopathy, the interventricular septum for hypertrophic cardiomyopathy, and more variable patterns for prior myocardial infarction). The performance of this model was comparable to that of the consensus of three senior cardiologists. These results also demonstrate how AI-assisted echocardiographic video image analysis enhances the accuracy of disease diagnostic classification [23].

6. Intracardiac Masses

Correct echocardiographic diagnosis of the etiology of intracardiac masses can be challenging but highly important, as different treatment options are possible for diverse types of cardiac masses (thrombosis, tumors, or vegetation), and this often requires further upstream testing with advanced imaging, such as MRI, for further characterization. AI technology could be applied to classify and recognize intracardiac masses, and previous research presented classification and segmentation results of intracardiac masses in echocardiograms using texture analysis [24]. This analysis was able to reflect some physiological properties of analyzed heart tissues. A more recent study investigated whether transesophageal echocardiography assisted with a computer-aided diagnostic algorithm was superior to the conventional approach in diagnosing left atrial thrombi in patients with atrial fibrillation [25]. The AI-derived algorithm significantly improved the diagnostic accuracy for left atrium thrombi when compared with the traditional approach by experts.

This entry is adapted from the peer-reviewed paper 10.3390/jimaging9020050

References

  1. Nedadur, R.; Wang, B.; Tsang, W. Artificial intelligence for the echocardiographic assessment of valvular heart disease. Heart 2022, 108, 1592–1599.
  2. Yoon, Y.E.; Kim, S.; Chang, H.J. Artificial Intelligence and Echocardiography. J. Cardiovasc. Imaging 2021, 29, 193–204.
  3. 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.
  4. Vafaeezadeh, M.; Behnam, H.; Hosseinsabet, A.; Gifani, P. A deep learning approach for the automatic recognition of prosthetic mitral valve in echocardiographic images. Comput. Biol. Med. 2021, 133, 104388.
  5. Martins, J.; Nascimento, E.R.; Nascimento, B.R.; Sable, C.A.; Beaton, A.Z.; Ribeiro, A.L.; Meira, W.; Pappa, G.L. Towards automatic diagnosis of rheumatic heart disease on echocardiographic exams through video-based deep learning. J. Am. Med. Inform. Assoc. 2021, 28, 1834–1842.
  6. Sengupta, P.P.; Shrestha, S.; Kagiyama, N.; Hamirani, Y.; Kulkarni, H.; Yanamala, N.; Bing, R.; Chin, C.W.L.; Pawade, T.A.; Messika-Zeitoun, D.; et al. A Machine-Learning Framework to Identify Distinct Phenotypes of Aortic Stenosis Severity. JACC Cardiovasc. Imaging 2021, 14, 1707–1720.
  7. Yang, F.; Chen, X.; Lin, X.; Chen, X.; Wang, W.; Liu, B.; Li, Y.; Pu, H.; Zhang, L.; Huang, D.; et al. Automated Analysis of Doppler Echocardiographic Videos as a Screening Tool for Valvular Heart Diseases. JACC Cardiovasc. Imaging 2022, 15, 551–563.
  8. Saraste, A.; Barbato, E.; Capodanno, D.; Edvardsen, T.; Prescott, E.; Achenbach, S.; Bax, J.J.; Wijns, W.; Knuuti, J. Imaging in ESC clinical guidelines: Chronic coronary syndromes. Eur Heart J. Cardiovasc. Imaging 2019, 20, 1187–1197.
  9. Paetsch, I.; Jahnke, C.; Ferrari, V.A.; Rademakers, F.E.; Pellikka, P.A.; Hundley, W.G.; Poldermans, D.; Bax, J.J.; Wegscheider, K.; Fleck, E.; et al. Determination of interobserver variability for identifying inducible left ventricular wall motion abnormalities during dobutamine stress magnetic resonance imaging. Eur. Heart J. 2006, 27, 1459–1464.
  10. Upton, R.; Mumith, A.; Beqiri, A.; Parker, A.; Hawkes, W.; Gao, S.; Porumb, M.; Sarwar, R.; Marques, P.; Markham, D.; et al. Automated Echocardiographic Detection of Severe Coronary Artery Disease Using Artificial Intelligence. JACC Cardiovasc. Imaging 2022, 15, 715–727.
  11. Pellikka, P.A. Artificially Intelligent Interpretation of Stress Echocardiography: The Future Is Now. JACC Cardiovasc. Imaging 2022, 15, 728–730.
  12. Laumer, F.; Di Vece, D.; Cammann, V.L.; Wurdinger, M.; Petkova, V.; Schonberger, M.; Schonberger, A.; Mercier, J.C.; Niederseer, D.; Seifert, B.; et al. Assessment of Artificial Intelligence in Echocardiography Diagnostics in Differentiating Takotsubo Syndrome from Myocardial Infarction. JAMA Cardiol. 2022, 7, 494–503.
  13. Zhou, J.; Du, M.; Chang, S.; Chen, Z. Artificial intelligence in echocardiography: Detection, functional evaluation, and disease diagnosis. Cardiovasc. Ultrasound 2021, 19, 29.
  14. Karakus, G.; Degirmencioglu, A.; Nanda, N.C. Artificial intelligence in echocardiography: Review and limitations including epistemological concerns. Echocardiography 2022, 39, 1044–1053.
  15. Yu, F.; Huang, H.; Yu, Q.; Ma, Y.; Zhang, Q.; Zhang, B. Artificial intelligence-based myocardial texture analysis in etiological differentiation of left ventricular hypertrophy. Ann. Transl. Med. 2021, 9, 108.
  16. Narula, S.; Shameer, K.; Salem Omar, A.M.; Dudley, J.T.; Sengupta, P.P. Machine-Learning Algorithms to Automate Morphological and Functional Assessments in 2D Echocardiography. J. Am. Coll. Cardiol. 2016, 68, 2287–2295.
  17. Goto, S.; Mahara, K.; Beussink-Nelson, L.; Ikura, H.; Katsumata, Y.; Endo, J.; Gaggin, H.K.; Shah, S.J.; Itabashi, Y.; MacRae, C.A.; et al. Artificial intelligence-enabled fully automated detection of cardiac amyloidosis using electrocardiograms and echocardiograms. Nat. Commun. 2021, 12, 2726.
  18. Duffy, G.; Cheng, P.P.; Yuan, N.; He, B.; Kwan, A.C.; Shun-Shin, M.J.; Alexander, K.M.; Ebinger, J.; Lungren, M.P.; Rader, F.; et al. High-Throughput Precision Phenotyping of Left Ventricular Hypertrophy with Cardiovascular Deep Learning. JAMA Cardiol. 2022, 7, 386–395.
  19. Hwang, I.C.; Choi, D.; Choi, Y.J.; Ju, L.; Kim, M.; Hong, J.E.; Lee, H.J.; Yoon, Y.E.; Park, J.B.; Lee, S.P.; et al. Differential diagnosis of common etiologies of left ventricular hypertrophy using a hybrid CNN-LSTM model. Sci. Rep. 2022, 12, 20998.
  20. Mahmood, R.; Syeda-Mahmood, T. Automatic detection of dilated cardiomyopathy in cardiac ultrasound videos. AMIA Annu. Symp. Proc. 2014, 2014, 865–871.
  21. Katsushika, S.; Kodera, S.; Nakamoto, M.; Ninomiya, K.; Kakuda, N.; Shinohara, H.; Matsuoka, R.; Ieki, H.; Uehara, M.; Higashikuni, Y.; et al. Deep Learning Algorithm to Detect Cardiac Sarcoidosis from Echocardiographic Movies. Circ. J. 2021, 86, 87–95.
  22. Sengupta, P.P.; Huang, Y.M.; Bansal, M.; Ashrafi, A.; Fisher, M.; Shameer, K.; Gall, W.; Dudley, J.T. Cognitive Machine-Learning Algorithm for Cardiac Imaging: A Pilot Study for Differentiating Constrictive Pericarditis from Restrictive Cardiomyopathy. Circ. Cardiovasc. Imaging 2016, 9, e004330.
  23. Liu, B.; Chang, H.; Yang, D.; Yang, F.; Wang, Q.; Deng, Y.; Li, L.; Lv, W.; Zhang, B.; Yu, L.; et al. A deep learning framework assisted echocardiography with diagnosis, lesion localization, phenogrouping heterogeneous disease, and anomaly detection. Sci. Rep. 2023, 13, 3.
  24. Strzelecki, M.; Materka, A.; Drozdz, J.; Krzeminska-Pakula, M.; Kasprzak, J.D. Classification and segmentation of intracardiac masses in cardiac tumor echocardiograms. Comput. Med. Imaging Graph. 2006, 30, 95–107.
  25. Sun, L.; Li, Y.; Zhang, Y.T.; Shen, J.X.; Xue, F.H.; Cheng, H.D.; Qu, X.F. A computer-aided diagnostic algorithm improves the accuracy of transesophageal echocardiography for left atrial thrombi: A single-center prospective study. J. Ultrasound Med. 2014, 33, 83–91.
More
© Text is available under the terms and conditions of the Creative Commons Attribution (CC BY) license; additional terms may apply. By using this site, you agree to the Terms and Conditions and Privacy Policy.
This entry is offline, you can click here to edit this entry!
Video Production Service
Feedback