Guson Kang

All Publications

• "A Responsible Framework for Applying Artificial Intelligence on Medical Images and Signals at the Point of Care: The PACS-AI Platform [Canadian Journal of Cardiology Volume 40, Issue 10, October 2024, Pages 1828-1840]". The Canadian journal of cardiology Theriault-Lauzier, P., Corbin, D., Tastet, O., Langlais, E. L., Taji, B., Kang, G., Chong, A., So, D., Tang, A., Gichoya, J. W., Chandar, S., Deziel, P., Hussin, J. G., Kadoury, S., Avram, R. 2025

  View details for DOI 10.1016/j.cjca.2025.08.203

  View details for PubMedID 41109328

• A deep learning phenome wide association study of the electrocardiogram. European heart journal. Digital health Hughes, J. W., Theurer, J., Vukadinovic, M., Rogers, A. J., Somani, S., Kang, G., Ghazizadeh, Z., O'Sullivan, J. W., Jain, S. S., Gomes, B., Salerno, M., Ashley, E., Zou, J. Y., Perez, M. V., Ouyang, D. 2025; 6 (4): 595-607

  Abstract

  Deep learning methods have shown impressive performance in detecting a range of diseases from electrocardiogram (ECG) waveforms, but the breadth of diseases that can be detected with high accuracy remains unknown, and in many cases the changes to the ECG allowing these classifications are also opaque. In this study, we aim to determine the full set of cardiac and non-cardiac conditions detectable from the ECG and to understand which ECG features contribute to the disease classification.Using large datasets of ECGs and connected electronic health records from two separate medical centres, we independently trained PheWASNet, a multi-task deep learning model, to detect 1243 different disease phenotypes from the raw ECG waveform. We confirmed that the ECG can be used to detect chronic kidney disease (AUC = 0.80), cirrhosis (AUC = 0.80), and sepsis (AUC = 0.84), as well as a range of cardiac diseases, and also found new detectable conditions, including respiratory failure (AUC = 0.86), neutropenia (AUC = 0.83), and menstrual disorders (AUC = 0.84). We found that of the 37 non-cardiac strongly detectable conditions, 35 were detectable by the model output for just four diseases, suggesting that they have similar effects on the ECG. We found that high performance in some conditions including neutropenia, respiratory failure, and sepsis can be explained by linear models based on conventional measurements taken from the ECG.Our study uncovers a range of diseases detectable in the ECG, including many previously unknown phenotypes, and makes progress towards understanding ECG features that allow this detection.

  View details for DOI 10.1093/ehjdh/ztaf047

  View details for PubMedID 40703109

  View details for PubMedCentralID PMC12282379

• A deep learning phenome wide association study of the electrocardiogram EUROPEAN HEART JOURNAL - DIGITAL HEALTH Hughes, J., Theurer, J., Vukadinovic, M., Rogers, A. J., Somani, S., Kang, G., Ghazizadeh, Z., O'Sullivan, J. W., Jain, S. S., Gomes, B., Salerno, M., Ashley, E., Zou, J. Y., Perez, M., Ouyang, D. 2025

  View details for DOI 10.1093/ehjdh/ztaf047

  View details for Web of Science ID 001490909200001

• Transcatheter Aortic Valve Replacement in Patients With Interventricular Membranous Septal Aneurysms. JACC. Case reports Theriault-Lauzier, P., Dawson, L. P., Wong, C. C., Sharma, R. P., Fearon, W. F., Kang, G., MacArthur, J. W., Pogatchnik, B. P., Krishnam, M. S., Yeung, A. C. 2025; 30 (1): 102715

  Abstract

  Aneurysms of the interventricular membranous septum are a rare anatomical feature that can be detected incidentally on computed tomography or echocardiography. Such aneurysms can pose challenges in the treatment of patients with aortic valve stenosis. A case series of 2 patients with membranous septal aneurysms treated successfully with current-generation balloon-expandable and self-expanding transcatheter heart valves is presented here. Preprocedural, intraprocedural, and postprocedural echocardiographic, fluoroscopic, and computed tomographic images are presented to demonstrate the feasibility of transcatheter aortic valve replacement in this patient population. The implications for device sizing and deployment dynamics are also illustrated.

  View details for DOI 10.1016/j.jaccas.2024.102715

  View details for PubMedID 39822800

  View details for PubMedCentralID PMC11733575

• Racial Inequities in Cardiovascular Procedure Use Influenced by Health- Related Social Needs Amponsah, D., Beyene, T., Song, N., Pillai, B., Ndukwe, M., Kang, G., Fearon, W., Asch, S., Heidenreich, P., Yong, C. ELSEVIER SCIENCE INC. 2024: B313-B314

  View details for Web of Science ID 001345355601310

• Racial and Ethnic Differences in Adoption of Mitral Valve Transcatheter Edge-to-Edge Repair Over a Decade in the National Veterans Affairs Healthcare System. Journal of the American Heart Association Pillai, B., Beyene, T. J., Kang, G., Amponsah, D., Heidenreich, P. A., Yong, C. M. 2024; 13 (19): e035767

  View details for DOI 10.1161/JAHA.124.035767

  View details for PubMedID 39344644

• A responsible framework for applying artificial intelligence on medical images and signals at the point-of-care: the PACS-AI platform. The Canadian journal of cardiology Theriault-Lauzier, P., Cobin, D., Tastet, O., Langlais, E. L., Taji, B., Kang, G., Chong, A., So, D., Tang, A., Gichoya, J. W., Chandar, S., Deziel, P., Hussin, J. G., Kadoury, S., Avram, R. 2024

  Abstract

  The potential of artificial intelligence (AI) in medicine lies in its ability to enhance clinicians' capacity to analyze medical images, thereby improving diagnostic precision and accuracy, thus enhancing current tests. However, the integration of AI within healthcare is fraught with difficulties. Heterogeneity among healthcare system applications, reliance on proprietary closed-source software, and rising cyber-security threats pose significant challenges. Moreover, prior to their deployment in clinical settings, AI models must demonstrate their effectiveness across a wide range of scenarios and must be validated by prospective studies, but doing so requires testing in an environment mirroring the clinical workflow which is difficult to achieve without dedicated software. Finally, the use of AI techniques in healthcare raises significant legal and ethical issues, such as the protection of patient privacy, the prevention of bias, and the monitoring of the device's safety and effectiveness for regulatory compliance. This review describes challenges to AI integration in healthcare and provides guidelines on how to move forward. We describe an open-source solution that we developed which integrates AI models into the Picture Archives Communication System (PACS), called PACS-AI. This approach aims to increase the evaluation of AI models by facilitating their integration and validation with existing medical imaging databases. PACS-AI may overcome many current barriers to AI deployment and offers a pathway towards responsible, fair, and effective deployment of AI models in healthcare. Additionally, we propose a list of criteria and guidelines that AI researchers should adopt when publishing a medical AI model, to enhance standardization and reproducibility.

  View details for DOI 10.1016/j.cjca.2024.05.025

  View details for PubMedID 38885787

• Race, Sex, and Age Disparities in the Performance of ECG Deep Learning Models Predicting Heart Failure. Circulation. Heart failure Kaur, D., Hughes, J. W., Rogers, A. J., Kang, G., Narayan, S. M., Ashley, E. A., Perez, M. V. 2023: e010879

  Abstract

  Deep learning models may combat widening racial disparities in heart failure outcomes through early identification of individuals at high risk. However, demographic biases in the performance of these models have not been well-studied.This retrospective analysis used 12-lead ECGs taken between 2008 and 2018 from 326 518 patient encounters referred for standard clinical indications to Stanford Hospital. The primary model was a convolutional neural network model trained to predict incident heart failure within 5 years. Biases were evaluated on the testing set (160 312 ECGs) using the area under the receiver operating characteristic curve, stratified across the protected attributes of race, ethnicity, age, and sex.There were 59 817 cases of incident heart failure observed within 5 years of ECG collection. The performance of the primary model declined with age. There were no significant differences observed between racial groups overall. However, the primary model performed significantly worse in Black patients aged 0 to 40 years compared with all other racial groups in this age group, with differences most pronounced among young Black women. Disparities in model performance did not improve with the integration of race, ethnicity, sex, and age into model architecture, by training separate models for each racial group, or by providing the model with a data set of equal racial representation. Using probability thresholds individualized for race, age, and sex offered substantial improvements in F1 scores.The biases found in this study warrant caution against perpetuating disparities through the development of machine learning tools for the prognosis and management of heart failure. Customizing the application of these models by using probability thresholds individualized by race, ethnicity, age, and sex may offer an avenue to mitigate existing algorithmic disparities.

  View details for DOI 10.1161/CIRCHEARTFAILURE.123.010879

  View details for PubMedID 38126168

• A deep learning-based electrocardiogram risk score for long term cardiovascular death and disease. NPJ digital medicine Hughes, J. W., Tooley, J., Torres Soto, J., Ostropolets, A., Poterucha, T., Christensen, M. K., Yuan, N., Ehlert, B., Kaur, D., Kang, G., Rogers, A., Narayan, S., Elias, P., Ouyang, D., Ashley, E., Zou, J., Perez, M. V. 2023; 6 (1): 169

  Abstract

  The electrocardiogram (ECG) is the most frequently performed cardiovascular diagnostic test, but it is unclear how much information resting ECGs contain about long term cardiovascular risk. Here we report that a deep convolutional neural network can accurately predict the long-term risk of cardiovascular mortality and disease based on a resting ECG alone. Using a large dataset of resting 12-lead ECGs collected at Stanford University Medical Center, we developed SEER, the Stanford Estimator of Electrocardiogram Risk. SEER predicts 5-year cardiovascular mortality with an area under the receiver operator characteristic curve (AUC) of 0.83 in a held-out test set at Stanford, and with AUCs of 0.78 and 0.83 respectively when independently evaluated at Cedars-Sinai Medical Center and Columbia University Irving Medical Center. SEER predicts 5-year atherosclerotic disease (ASCVD) with an AUC of 0.67, similar to the Pooled Cohort Equations for ASCVD Risk, while being only modestly correlated. When used in conjunction with the Pooled Cohort Equations, SEER accurately reclassified 16% of patients from low to moderate risk, uncovering a group with an actual average 9.9% 10-year ASCVD risk who would not have otherwise been indicated for statin therapy. SEER can also predict several other cardiovascular conditions such as heart failure and atrial fibrillation. Using only lead I of the ECG it predicts 5-year cardiovascular mortality with an AUC of 0.80. SEER, used alongside the Pooled Cohort Equations and other risk tools, can substantially improve cardiovascular risk stratification and aid in medical decision making.

  View details for DOI 10.1038/s41746-023-00916-6

  View details for PubMedID 37700032

  View details for PubMedCentralID 8145781

• Distance between valvular leaflet and coronary ostium predicting risk of coronary obstruction during TAVR. International journal of cardiology. Heart & vasculature Oh, J., Kobayashi, Y., Kang, G., Nishi, T., Willemink, M. J., Fearon, W. F., Fischbein, M., Fleishmann, D., Yeung, A. C., Kim, J. B. 1800; 37: 100917

  Abstract

  Background: The aim of this study was to evaluate the role of the distance between the aortic valve in projected position to the coronary ostium to determine risk of coronary artery obstruction after transcatheter aortic valve replacement (TAVR).Methods: An Expected Leaflet-to-ostium Distance (ELOD) was obtained on pre-TAVR planning computed tomography by subtracting leaflet thickness and the distances from the center to the annular rim at annulus level and from the center to the coronary ostium at mid-ostial level. Variables were compared between patients with and without coronary obstruction and the level of association between variables was assessed using log odds ratio (OR).Results: A total of 177 patients with 353 coronary arteries was analyzed. Mean annulus diameters (22.8±2.8mm and 23.4±1.0mm, p>0.05) and mean sinus of Valsalva (SOV) diameters (31.2±3.6mm and 31.9±3.6mm, p>0.05) were similar between patients with lower and higher coronary heights, respectively. There were three coronary obstruction cases. ELOD≤2mm in combination with leaflet length longer than mid-ostial height allowed for discrimination of cases with and without coronary obstruction. There was a significant association between coronary obstruction event and ELOD≤2mm (log OR=6.180, p<0.001).Conclusions: Our study showed that a combination of ELOD<2mm and a longer leaflet length than mid-ostial height may be associated with increased risk for coronary obstruction during TAVR.

  View details for DOI 10.1016/j.ijcha.2021.100917

  View details for PubMedID 34917750

• Choosing Between Transcatheter Aortic Valve Replacement and Surgery in the Low-Risk Transcatheter Aortic Valve Replacement Era. Interventional cardiology clinics Kang, G., Yeung, A. 2021; 10 (4): 413-422

  Abstract

  The landmark results of the low surgical risk pivotal transcatheter aortic valve replacement (TAVR) trials fueled speculation that the role of surgical aortic valve replacement (SAVR) would be limited in the future. Instead, the field has pivoted away from reductive surgical risk stratification toward understanding the complex interplay of anatomy, timing, and surgical risk to optimize the lifetime management of aortic stenosis. In this review, we systematically explore the subtleties that influence the choice between TAVR and surgery in the low-risk TAVR era.

  View details for DOI 10.1016/j.iccl.2021.05.001

  View details for PubMedID 34593105

• Chase the Leak - A Case of Valve-in-Ring with Mitral PVL Closure So, C., Kang, G., Eng, M. ELSEVIER SCIENCE INC. 2021: S247–S248

  View details for Web of Science ID 000637884100174

• Deep Neural Network Trained on Surface ECG Improves Diagnostic Accuracy of Prior Myocardial Infarction Over Q Wave Analysis Yildirim, O., Baloglu, U. B., Talo, M., Ganesan, P., Tung, J. S., Kang, G., Tooley, J., Alhusseini, M., Baykaner, T., Wang, P. J., Perez, M., Tereshchenko, L., Narayan, S. M., Rogers, A. J., IEEE IEEE. 2021

  View details for DOI 10.22489/CinC.2021.010

  View details for Web of Science ID 000821955000130

• Spontaneous Coronary Artery Dissection and ST-Segment Elevation Myocardial Infarction in an Anomalous LAD Artery JACC: Case Reports Kang, G., Sarraju, A., Nishi, T., Rogers, I., Tremmel, J., Kim, J. 2020

  View details for DOI 10.1016/j.jaccas.2019.11.061

• A novel noninvasive method for remote heart failure monitoring: the EuleriAn video Magnification apPLications In heart Failure studY (AMPLIFY). NPJ digital medicine Abnousi, F., Kang, G., Giacomini, J., Yeung, A., Zarafshar, S., Vesom, N., Ashley, E., Harrington, R., Yong, C. 2019; 2: 80

  Abstract

  Current remote monitoring devices for heart failure have been shown to reduce hospitalizations but are invasive and costly; accurate non-invasive options remain limited. The EuleriAn Video Magnification ApPLications In Heart Failure StudY (AMPLIFY) pilot aimed to evaluate the accuracy of a novel noninvasive method that uses Eulerian video magnification. Video recordings were performed on the neck veins of 50 patients who were scheduled for right heart catheterization at the Palo Alto VA Medical Center. The recorded jugular venous pulsations were then enhanced by applying Eulerian phase-based motion magnification. Assessment of jugular venous pressure was compared across three categories: (1) physicians who performed bedside exams, (2) physicians who reviewed both the amplified and unamplified videos, and (3) direct invasive measurement of right atrial pressure from right heart catheterization. Motion magnification reduced inaccuracy of the clinician assessment of central venous pressure compared to the gold standard of right heart catheterization (mean discrepancy of -0.80cm H2O; 95% CI -2.189 to 0.612, p=0.27) when compared to both unamplified video (-1.84cm H2O; 95% CI -3.22 to -0.46, p=0.0096) and the bedside exam (-2.90cm H2O; 95% CI -4.33 to 1.40, p=0.0002). Major categorical disagreements with right heart catheterization were significantly reduced with motion magnification (12%) when compared to unamplified video (25%) or the bedside exam (27%). This novel method of assessing jugular venous pressure improves the accuracy of the clinical exam and may enable accurate remote monitoring of heart failure patients with minimal patient risk.

  View details for DOI 10.1038/s41746-019-0159-0

  View details for PubMedID 31453375

• Expanding transcatheter aortic valve replacement into uncharted indications. The Korean journal of internal medicine Kang, G., Kim, J. B. 2018; 33 (3): 474–82

  Abstract

  Since the first-in-man transcatheter delivery of an aortic valve prosthesis in 2002, the landscape of aortic stenosis therapeutics has shifted dramatically. While initially restricted to non-surgical cases, progressive advances in transcatheter aortic valve replacement and our understanding of its safety and efficacy have expanded its use in intermediate and possibly low surgical risk patients. In this review, we explore the past, present, and future of transcatheter aortic valve replacement.

  View details for PubMedID 29551053

• Neprilysin Inhibitors in Cardiovascular Disease. Current cardiology reports Kang, G., Banerjee, D. 2017; 19 (2): 16-?

  Abstract

  Mortality from heart failure remains high despite advances in medical therapy over the last three decades. Angiotensin receptor-neprilysin inhibitor (ARNI) combinations are the latest addition to the heart failure medical armamentarium, which is built on the cornerstone regimen of beta blockers, angiotensin converting enzyme (ACE) inhibitors/angiotensin receptor blockers, and aldosterone antagonists. Recent trial data have shown a significant mortality benefit from ARNIs, which, as of May 2016, have now received a class I recommendation for use in patients with heart failure and reduced ejection fraction from the major American and European cardiology societies.

  View details for DOI 10.1007/s11886-017-0827-0

  View details for PubMedID 28185171

• Pulmonary artery pulsatility index predicts right ventricular failure after left ventricular assist device implantation. journal of heart and lung transplantation Kang, G., Ha, R., Banerjee, D. 2016; 35 (1): 67-73

  Abstract

  Right ventricular failure (RVF) is a major cause of morbidity and mortality after left ventricular assist device (LVAD) implantation. The pulmonary artery pulsatility index (PAPi) is a novel hemodynamic index that predicts RVF in the setting of myocardial infarction, although it has not been shown to predict RVF after LVAD implantation.We performed a retrospective, single-center analysis to examine the utility of the PAPi in predicting RVF and RV assist device (RVAD) implantation in 85 continuous-flow LVAD recipients. We performed a multivariate logistic regression analysis incorporating previously identified predictors of RVF after LVAD placement, including clinical and echocardiographic variables, to determine the independent effect of PAPi in predicting RVF or RVAD after LVAD placement.In this cohort, the mean PAPi was 3.4 with a standard deviation of 2.9. RVF occurred in 33% of patients, and 11% required a RVAD. Multivariate analysis, adjusting for age, blood urea nitrogen (BUN), and Interagency Registry for Mechanically Assisted Circulatory Support profile, revealed that higher PAPi was independently associated with a reduced risk of RVAD placement (odds ratio [OR], 0.30; 95% confidence interval [CI], 0.07-0.89). This relationship did not change significantly when echocardiographic measures were added to the analysis. Stratifying the analysis by the presence of inotropes during catheterization revealed that PAPi was more predictive of RVAD requirement when measured on inotropes (OR, 0.21; 95% CI, 0.02-0.97) than without (OR, 0.49; 95% CI, 0.01-1.94). Furthermore, time from catheterization to LVAD did not significantly affect the predictive value of the PAPi (maximum time, 6 months). Receiver operating characteristic curve analysis revealed that optimal sensitivity and specificity were achieved using a PAPi threshold of 2.0.In LVAD recipients, the PAPi is an independent predictor of RVF and the need for RVAD support after LVAD implantation. This index appears more predictive in patients receiving inotropes and was not affected by time from catheterization to LVAD in our cohort.

  View details for DOI 10.1016/j.healun.2015.06.009

  View details for PubMedID 26212656