Searched for: in-biosketch:true
person:jankel01
Deep learning models for electrocardiograms are susceptible to adversarial attack
Han, Xintian; Hu, Yuxuan; Foschini, Luca; Chinitz, Larry; Jankelson, Lior; Ranganath, Rajesh
Electrocardiogram (ECG) acquisition is increasingly widespread in medical and commercial devices, necessitating the development of automated interpretation strategies. Recently, deep neural networks have been used to automatically analyze ECG tracings and outperform physicians in detecting certain rhythm irregularities1. However, deep learning classifiers are susceptible to adversarial examples, which are created from raw data to fool the classifier such that it assigns the example to the wrong class, but which are undetectable to the human eye2,3. Adversarial examples have also been created for medical-related tasks4,5. However, traditional attack methods to create adversarial examples do not extend directly to ECG signals, as such methods introduce square-wave artefacts that are not physiologically plausible. Here we develop a method to construct smoothed adversarial examples for ECG tracings that are invisible to human expert evaluation and show that a deep learning model for arrhythmia detection from single-lead ECG6 is vulnerable to this type of attack. Moreover, we provide a general technique for collating and perturbing known adversarial examples to create multiple new ones. The susceptibility of deep learning ECG algorithms to adversarial misclassification implies that care should be taken when evaluating these models on ECGs that may have been altered, particularly when incentives for causing misclassification exist.
PMID: 32152582
ISSN: 1546-170x
CID: 4349692
Esophageal Temperature Dynamics During High Power Short Duration Posterior Wall Ablation
Barbhaiya, Chirag R; Kogan, Edward V; Jankelson, Lior; Knotts, Robert J; Spinelli, Michael; Bernstein, Scott; Park, David; Aizer, Anthony; Chinitz, Larry A; Holmes, Douglas
BACKGROUND:Increased peak luminal esophageal temperature (LET) is associated with increased risk of esophageal injury following left atrial posterior wall (LAPW) ablation. The magnitude, distribution, and risk factors of LET increase with high power short duration (HPSD) LAPW ablation are not well understood. OBJECTIVE:We aimed to describe the spatial and temporal characteristics of LET changes associated with HPSD LAPW RFA. METHODS:LET was sampled at 20Hz using a 12-point esophageal temperature monitor (CIRCA S-CATH, Circa Scientific, Inc.) in 16 patients undergoing LAPW ablation. Esophageal temperature sensor position and lesion locations were recorded using an electroanatomic mapping system with fluoroscopic integration (CARTO 3, CARTOUNIVU, Biosense Webster, Inc). Point-by-point LAPW ablation was performed at 50W for 6s. The first 20 LAPW lesions were individually analyzed in each patient. RESULTS:LET increase ≥4°C (8 lesions: Max LET 5.8°C), 2-4°C (34 lesions), and 1-2°C (58 lesions) occurred at 9±2 mm, 8±2 mm, and 13±2mm from sensors, respectively. Lesions placed >20mm from a temperature sensor did not result in an LET increase ≥2°C. Temperature resolution to within 1°C of baseline occurred at ∼60s after cessation of RF. Consecutive lesions resulting in additive heating of at least 1°C occurred in 17 lesion pairs with an inter-lesion distance of 9±4mm and inter-lesion time of 21±4s. CONCLUSION/CONCLUSIONS:HPSD LAPW ablation can result in severe esophageal temperature increases. Significant LET increase will be undetected when lesions are >20mm away from a temperature sensor. Additive LET increase was observed with consecutive lesions placed less than 20mm apart within 60s.
PMID: 31978595
ISSN: 1556-3871
CID: 4273622
Quantitative analysis of ablation technique predicts arrhythmia recurrence following atrial fibrillation ablation
Jankelson, Lior; Dai, Matthew; Bernstein, Scott; Park, David; Holmes, Douglas; Aizer, Anthony; Chinitz, Larry; Barbhaiya, Chirag
BACKGROUND:Optimal ablation technique, including catheter-tissue contact during atrial fibrillation (AF) radiofrequency (RF) ablation, is associated with improved procedural outcomes. We used a custom developed software to analyze high-frequency catheter position data to study the interaction between catheter excursion during lesion placement, lesion-set sequentiality, and arrhythmia recurrence. METHODS:A total of 100 consecutive patients undergoing first-time RF ablation for paroxysmal AF were analyzed. Spatial positioning of the ablation catheter sampled at 60 Hz during RF application was extracted from the CARTO3 system (Biosense Webster Inc, USA) and analyzed using custom-developed MATLAB software to determine precise catheter spatial 3D excursion during RF ablation. The primary end point was freedom from atrial arrhythmia lasting longer than 30 seconds after a single ablation procedure. RESULTS:At 1 year, 86% of patients were free from recurrent arrhythmia. There was no significant difference in clinical, echocardiographic, or ablation characteristics between patients with and without recurrent arrhythmia. Analyzing 15,356,998 position data points revealed that lesion-set sequentiality and mean lesion catheter excursion were predictors of arrhythmia recurrence. Analyzing arrhythmia recurrence by mean single-lesion catheter excursion (excursion >2.81 mm) and by sequentiality (using 46% of lesions with interlesion distance >6 mm as cutoff) revealed significantly increased arrhythmia recurrence in the higher excursion group (23% vs 6%, P = .03) and in the less sequential group (24% vs 4%, P = .02). CONCLUSIONS:Ablation lesion sequentiality measured by catheter interlesion distance and catheter stability measured by catheter excursion during lesion placement are potentially modifiable factors affecting arrhythmia recurrence after RF ablation for AF.
PMID: 31835167
ISSN: 1097-6744
CID: 4235072
Impact of RNA testing on cardiac variant interpretation and patient management [Case Report]
Jankelson, Lior; Zimmermann, Heather; Fowler, Steven; Borneman, Linda; Conner, Blair; Wu, Sitao; Karam, Rachid; Chinitz, Larry; Cerrone, Marina
PMCID:6700669
PMID: 31453089
ISSN: 2214-0271
CID: 4054342
PERSISTENT ATRIAL FIBRILLATION CATHETER ABLATION OUTCOMES STRATIFIED BY LEFT ATRIAL POSTERIOR WALL ISOLATION LESION SET CHARACTERISTICS [Meeting Abstract]
Barbhaiya, C R; Kogan, E V; Knotts, R; Pelaez, A V; Jankelson, L; Bernstein, S A; Park, D S; Holmes, D; Aizer, A; Chinitz, L A
Background: Left atrial posterior wall (LAPW) isolation is associated with favorable outcomes for catheter ablation of persistent atrial fibrillation (AF) in several studies. Reported techniques for LAPW isolation include ablation at the periphery without ablation within the LAPW, and high density ablation of all sites of electrical activity within the LAPW. The proportion of LA isolated by the lesion set in various reports also varies greatly. The optimal technique to achieve LAPW isolation is not clear.
Objective(s): To assess impact of ablation lesion density within the LAPW and dimensions of LAPW isolation region on arrhythmia recurrence in catheter ablation of persistent AF.
Method(s): LAPW lesion density, and LAPW isolation surface area relative to total LA surface area were calculated using electroanatomic maps of 110 consecutive patients undergoing LAPW isolation for persistent AF (CARTO 3, Biosense Webster, Inc.) LAPW isolation lesion sets were created at the discretion of 5 experienced operators after LA voltage mapping. LAPW and PV entrance block and exit block were confirmed. Arrhythmia recurrence at one year was assessed by the Kaplan-Meier method.
Result(s): LAPW lesion density ranged from 0% - 99%. Proportion of LA surface area isolated ranged from 35% - 75%. There was no significant difference in arrhythmia-free survival by quartile of LAPW ablation density (81% vs. 68% vs. 85% vs. 78%, p=0.8), or by quartile of LA surface area proportion isolated (85% vs. 75% vs. 79% vs. 74%, p=0.3). Voltage map guided LAPW isolation resulted in no significant difference in incidence of recurrent arrhythmia by quartile of total LA surface area (81% vs. 78% vs. 78% vs. 74%, p=0.5).
Conclusion(s): Neither the density of ablation within the LAPW nor the dimensions of the isolated region predicted arrhythmia-free survival LAPW isolation for catheter ablation of persistent AF. Voltage map guided LAPW isolation resulted in similar ablation efficacy regardless of LA size.
Copyright
EMBASE:2002272949
ISSN: 1556-3871
CID: 4007262
PSEUDOPOLYMORPHIC WIDE COMPLEX TACHYCARDIA IN A CHILD WITH LQT2 [Meeting Abstract]
Jankelson, L; Magnani, S; Cecchin, F; Tan, R; Barbhaiya, C R; Aizer, A; Holmes, D; Bernstein, S A; Park, D S; Borneman, L; Cerrone, M; Chinitz, L A
Background: Implantable loop recorder (ILR) based monitoring of patients with LQTS allows enhanced arrhythmia surveillance and can help distinguish life-threatening from benign arrhythmias.
Objective(s): We present a case of a child with LQTS and wide complex tachycardia detected by ILR.
Result(s): An asymptomatic 12 year old with LQT2 syndrome, positive for a G648S hERG mutation, with baseline QTc of 510-550ms despite maximally tolerated Nadolol (Figure 1A) was followed in our inherited arrhythmia center. His affected mother has had multiple syncopal events related to polymorphic ventricular tachycardia (VT) and appropriate ICD shocks. We elected to implant him with ILR to allow longitudinal monitoring and early detection of arrhythmia. He presented 6 months later with 2 alerts for asymptomatic polymorphic, wide complex tachycardia at ~200 bpm during sleeping (Figure 1B). Electrophysiology study (EPS) was performed to determine etiology of the arrhythmia. Dual AV node physiology was present. Sinus tachycardia at 200 bpm with left bundle branch block (LBBB) morphology was induced with Isoproterenol and atrio-fascicular pathway was excluded. Respiratory changes resulted in the tachycardia appearing as polymorphic on the ILR during the EPS.
Conclusion(s): This is the first reported case of sinus tachycardia with LBBB aberrancy in a child with LQTS. Pseudopolymorphic wide complex tachycardia was the result of aberrancy and respiratory artifact. Combined ILR monitoring and EP study provided a correct diagnosis, thus avoiding further interventions. [Figure presented]
Copyright
EMBASE:2002272661
ISSN: 1556-3871
CID: 4007282
POST ELECTROPHYSIOLOGY PROCEDURE MONITORING WITH A NOVEL WEARABLE DEVICE [Meeting Abstract]
Jankelson, L; Bennet, M; Barbhaiya, C R; Aizer, A; Holmes, D; Bernstein, S A; Park, D S; Chinitz, L A
Background: The rapid increase in the number and complexity of electrophysiology (EP) procedures performed annually requires improved strategies to safely enhance post procedural monitoring and early discharge.
Objective(s): To determine if a wearable device that continuously and non-invasively measures vital signs and hemodynamic parameters can predict post-procedure cardiovascular decompensation.
Method(s): Investigator initiated, prospective study of 21 patients receiving either atrial fibrillation ablation (N=16) or device implantation (N=5). Patients were monitored for the post-procedure time periods ranging from 8 to18 hours with two techniques: 1) sporadic measurements with a conventional vital sign monitor; and 2) continuous measurements with toSense's CoVaTM Monitoring System, an FDA-cleared wearable device measuring stroke volume (SV), cardiac output (CO), thoracic fluid index (TFI), heart rate (HR), and respiration rate (RR). CoVaTM-generated data were wirelessly analyzed to established markers of decompensation, defined as decline in systolic blood pressure of >20mmHg over 5 minutes or SpO2 values < 90%. Decompensation index (DI), a normalized index defined as: DI = mean{(norm d[HR]/dt) + (norm d[HRV]/dt) + (norm d[SV]/dt) + (norm d[RR]/dt) + (norm [TI]/dt)} was used to predict decompensation events.
Result(s): Patients continuously monitored for >8 hours with CoVaTM, who were also sporadically measured at least 5 times with the vital sign monitor (N=11) were analyzed. Agreement of the two parameters measured by both devices-HR and RR-was: DELTAHR = 6.5+/-0.3 bpm; DELTARR = 3.8+/-0.8 breaths/min. 55% of patients (N=6) experienced brief decompensation events. The vital sign monitor made sporadic measurements every 95+/-21.3 minutes. Using DI, CoVaTM predicted 78% of the total patient decompensations, with the average prediction being 22+/-23 minutes in advance.
Conclusion(s): Continuous measurements with a wearable device may detect post-procedure decompensation in patients receiving electrophysiology procedures or device implantations with greater accuracy and better temporal resolution as compared to sporadic measurements with conventional vital sign monitors.
Copyright
EMBASE:2002272580
ISSN: 1556-3871
CID: 4007292
Non-invasive cardiac mapping for non-response in cardiac resynchronization therapy
Strik, Marc; Ploux, Sylvain; Jankelson, Lior; Bordachar, Pierre
Cardiac resynchronization therapy (CRT) is an effective intervention in selected patients with moderate-to-severe heart failure with reduced ejection fraction and abnormal left ventricular activation time. The non-response rate of approximately 30% has remained nearly unchanged since this therapy was introduced 25 years ago. While intracardiac mapping is widely used for diagnosis and guidance of therapy in patients with tachyarrhythmia, its application in characterization of the electrical substrate to elucidate the mechanisms involved in CRT response remain anecdotal. In the present review, we describe the traditional determinants of CRT response before presenting novel non-invasive techniques used for CRT optimization. We discuss efforts to identify the target electrical substrate to guide the deployment of pacing electrodes during the operative procedure. Non-invasive body surface mapping technologies such as ECG imaging or ECG belt enables prediction of acute and chronic CRT response. While electrical dyssynchrony parameters provide high predictive accuracy for CRT response when obtained during intrinsic conduction, their predictive value is less when acquired during CRT or LV-pacing. Key messages Classic predictors of CRT response are female gender, NYHA class ≤ III, left ventricular ejection fraction ≥25%, QRS duration ≥150 ms and estimated glomerular filtration rate ≥60 mL/min. ECG-imaging is a comprehensive non-invasive mapping system which allows to express the amount of electrical asynchrony of a CRT candidate. Non-invasive body surface mapping technologies enables excellent prediction of acute and chronic CRT response before implantation. When performed during CRT or LV-pacing, the added value of these mapping systems remains unclear.
PMID: 31094217
ISSN: 1365-2060
CID: 3999992
RAPID PACING AND HIGH FREQUENCY JET VENTILATION SYNERGISTICALLY IMPROVE CATHETER STABILITY DURING AF ABLATION [Meeting Abstract]
Aizer, A; Qiu, J K; Cheng, A; Wu, P; Holmes, D; Jankelson, L; Bernstein, S A; Park, D S; Linton, P; Barbhaiya, C R; Chinitz, L A
Background: Increased catheter stability during AF ablation is associated with higher ablation success rates. Rapid cardiac pacing and high frequency jet ventilation (HFJV) have both been independently shown to improve catheter stability. Simultaneous modulation of cardiac and respiratory motion has not been previously studied.
Objective(s): To study the effects of modulating heart rate and respiratory rate simultaneously on catheter stability.
Method(s): Forty paroxysmal AF patients were randomized to four study conditions. Ablation lesions were created at 15 prespecified locations. Twenty patients received atrial pacing (500 msec) during the first half of each lesion; twenty patients were paced during the second half of each lesion. Within each group, half received HFJV and half received standard ventilation. Contact force (CF) variability, defined as CF standard deviation, was compared between study groups.
Result(s): Compared to sinus rhythm and standard ventilation, rapid pacing (5.45 g vs. 5.86 g; p=0.006) and HFJV (5.10 g; p=0.003) each significantly reduced mean CF standard deviation. Simultaneous pacing and HFJV produced even greater reduction of mean CF standard deviation (4.29 g; p<0.001) (Figure). Pacing and HFJV alone had similar effects on mean CF variability (p=0.2).
Conclusion(s): Rapid pacing and HFJV synergistically improve catheter stability during AF ablation. Simultaneous pacing with HFJV further optimizes catheter stability over pacing or HFJV alone and may improve ablation outcomes. [Figure presented]
Copyright
EMBASE:2002296075
ISSN: 1556-3871
CID: 4001842
LONG-TERM ARRHYTHMIA RECURRENCE AFTER ATRIAL FIBRILLATION ABLATION IN HYPERTROPHIC CARDIOMYOPATHY [Meeting Abstract]
Jankelson, L; Kogan, E V; Barbhaiya, C R; Aizer, A; Holmes, D; Park, D S; Stepanovic, A; Cerrone, M; Sherrid, M; Chinitz, L A
Background: Despite the increased prevalence of atrial fibrillation (AF) in hypertrophic cardiomyopathy (HCM), the efficacy of radiofrequency ablation (RFA) has been characterized over limited follow-up intervals (~1 year). Several large meta-analyses note that patients with HCM have substantially higher rates of arrhythmia recurrence after RFA, compared to patients without HCM. The implication of confirmed HCM mutations on ablation efficacy has similarly only been assessed in small-scale studies.
Objective(s): To assess arrhythmia recurrence after RFA in patients with HCM and paroxysmal AF (PAAF) or persistent AF (PEAF) as well as its relation to their genetic background and LVOT gradient.
Method(s): Arrhythmia recurrence after RFA was assessed in 66 HCM patients and compared to 343 patients without HCM. AF recurrence was defined as AF on EKG or >30s of AF on ICD/pacemaker interrogation or on monitoring devices after a 3-month blanking period. Kaplan-Meier analysis was performed to compare arrhythmia recurrence rate and timing.
Result(s): The EF of HCM patients was higher than that of the non-HCM patients in both the PAAF and PEAF groups (65.5 and 63.0% vs 61.4 and 53.3%, respectively, p<0.001); within the HCM group, the clinical characteristics of the genetically (+) HCM group (n=14) did not differ from those of the genetically (-) group (n=12). Arrhythmia recurrence at 1 year in PAAF and PEAF was not significantly different between HCM and non-HCM patients (18% vs 11%, p=0.2, and 33% vs 26%, p=1), nor was mean time to arrhythmia recurrence (PAAF 193+/-48 vs 181+/-59 days, p=0.8, and PEAF 175+/-58 vs 168+/-20 days, p=0.6). Recurrence rates over the entire follow-up period of the HCM patients were 54 and 85% in the PAAF and PEAF groups (1076+/-187 and 1050+/- 201 days of follow-up), respectively. Amongst HCM patients with LVOT gradients >70mmHg (PAAF, n = 8, and PEAF, n = 3) longer-term rates of arrhythmia recurrence were similar at 88% and 67% (p=0.9).
Conclusion(s): Arrhythmia recurrence at 1 year following AF ablation in HCM patients is similar to that of non-HCM AF patients regardless of the type of AF. Absolute rates of atrial arrhythmia recurrence in HCM patients at >3 years post ablation are considerable. Confirmed HCM mutations and severe LVOT gradients do not modify the outcome of AF ablation.
Copyright
EMBASE:2002296056
ISSN: 1556-3871
CID: 4001852