How Do Cardiology Practitioners Interpret an Electrocardiogram: An Eye Tracking Study (Preprint)

BACKGROUND Visual expertise refers to advanced visual skills demonstrated when executing domain‐specific visual tasks. Understanding healthcare practitioners’ visual expertise across different levels in the healthcare sector is crucial in clarifying how to acquire accurate interpretations of electrocardiograms (ECGs). OBJECTIVE The study aims to quantify, through the use of eye-tracking, differences in the visual expertise of medical practitioners, such as medical students, cardiology nurses, technicians, fellows, and consultants, when interpreting ECGs. METHODS Sixty-three participants with different healthcare roles participated in an eye-tracking study that consisted of interpreting 10 ECGs with different heart abnormalities. A counterbalanced within-subjects design was employed with one independent variable consisting of the expertise level of the medical practitioners and two measured eye-tracking dependent variables (fixations count and fixations revisitation). Eye-tracking data was assessed according to the accuracy of interpretation and frequency interpreters visited different leads in ECGs. In addition, the median and standard deviation in the interquartile range for the fixations count and the mean and standard deviation for the ECG lead revisitations were calculated. RESULTS Accuracy of interpretation ranged between 98% among consultants and 52% among medical students. Eye-tracking features also reflected this difference in the accuracy of interpretation. The results of the eye fixations count and eye fixations revisitations indicate that the less experienced medical practitioners need to observe various ECG leads more carefully. However, experienced medical practitioners rely on visual pattern recognition to provide their ECG diagnoses. CONCLUSIONS The results show that visual expertise for ECG interpretation is linked to the practitioner’s role within the healthcare system and the number of years of practical experience interpreting ECGs. Medical practitioners focus on different ECG leads and different waveform abnormalities according to their role in the healthcare sector and their expertise levels.

Online educational module for paramedics on prehospital 15-lead ECG recording – results of the educational part of the “Finding LCX AMI With posterior ECG leads” (FLAWLESS) trial

Background The diagnosis of ST elevation myocardial infarction (STEMI) is challenging when the culprit is in the left circumflex coronary artery (CX) territory because ST elevations are often not captured by the standard 12-lead electrocardiogram (ECG). Although, guidelines recommend the acquisition of the additional posterior leads V7-V9 (pECG) when the suspicion of acute coronary syndrome (ACS) is high and the ECG non-diagnostic, this is not routinely done. Purpose The purpose of the FLAWLESS trial, was to improve the prehospital CX STEMI diagnostic. The study consisted of 2 parts: a) a training and implementation study, and b) an outcomes study after implementation. In the implementation study we evaluated the FLAWLESS process from the paramedic's point of view on experiences, implementation of pECG lead recordings and its barriers. Methods Before initiating the trial, all active paramedics in 2 health care regions were educated via a specifically designed and mandatory online 30 min course and all 250 ambulances equipped with a SMART-CARD (instructing how to record pECG leads) and FAQ-sheet. All paramedics were invited by email to anonymously answer an online questionnaire (OQ) designed in REDCap® and interviewed. Utility-score and difficulty-score, ranging from 0 (not useful at all/very easy) to 100 (very useful/very difficult), were introduced for quantitative assessments. Results A total of 1268 paramedics were invited to answer the OQ. The response rate was intermediate at 35%. Among responders, 89% had completed the OEP. On duty 80% had used FAQ-sheet and 74% SMART-CARD in the field. The median utility scores were 80 (25th and 75th quartiles 67–90) for OEP, 79 (61–90) for FAQ-sheet and 85 (75–97) for SMART-CARD, respectively. The implementation of pECG leads recordings was fairly high – 54% reported always recording V7-V9 in ACS patients and 36% reported doing it frequently. Difficulty-score for recording V7-V9 leads in the prehospital setting was 50 (19–70). Finally, 43% reported difficulties that were related to technicalities i.e. defibrillators not having dedicated V7, V8 and V9 cables, hence ambulance staff is forced to record and transmit a second ECG after moving the V4, V5 and V6 cables to the V7-V9 positioned electrodes. Conclusion We demonstrated that large-scale online training of paramedics in the recording of prehospital 15-lead ECG is feasible. The evaluation was positive regarding training and support tools in the ambulances but almost 50% of paramedics found the recording very difficult in the field. Future ECG machines used in emergency settings should be constructed with 13 instead of 10 cables to allow simultaneously recording of 15 leads (standard, precordial and the V7-V9 posterior). This would ease acquisition, facilitate implementation of guideline recommendation. FUNDunding Acknowledgement Type of funding sources: None.

Automated Detection of Acute Myocardial Infarction Using Asynchronous Electrocardiogram Signals—Preview of Implementing Artificial Intelligence With Multichannel Electrocardiographs Obtained From Smartwatches: Retrospective Study

Background When using a smartwatch to obtain electrocardiogram (ECG) signals from multiple leads, the device has to be placed on different parts of the body sequentially. The ECG signals measured from different leads are asynchronous. Artificial intelligence (AI) models for asynchronous ECG signals have barely been explored. Objective We aimed to develop an AI model for detecting acute myocardial infarction using asynchronous ECGs and compare its performance with that of the automatic ECG interpretations provided by a commercial ECG analysis software. We sought to evaluate the feasibility of implementing multiple lead–based AI-enabled ECG algorithms on smartwatches. Moreover, we aimed to determine the optimal number of leads for sufficient diagnostic power. Methods We extracted ECGs recorded within 24 hours from each visit to the emergency room of Ajou University Medical Center between June 1994 and January 2018 from patients aged 20 years or older. The ECGs were labeled on the basis of whether a diagnostic code corresponding to acute myocardial infarction was entered. We derived asynchronous ECG lead sets from standard 12-lead ECG reports and simulated a situation similar to the sequential recording of ECG leads via smartwatches. We constructed an AI model based on residual networks and self-attention mechanisms by randomly masking each lead channel during the training phase and then testing the model using various targeting lead sets with the remaining lead channels masked. Results The performance of lead sets with 3 or more leads compared favorably with that of the automatic ECG interpretations provided by a commercial ECG analysis software, with 8.1%-13.9% gain in sensitivity when the specificity was matched. Our results indicate that multiple lead-based AI-enabled ECG algorithms can be implemented on smartwatches. Model performance generally increased as the number of leads increased (12-lead sets: area under the receiver operating characteristic curve [AUROC] 0.880; 4-lead sets: AUROC 0.858, SD 0.008; 3-lead sets: AUROC 0.845, SD 0.011; 2-lead sets: AUROC 0.813, SD 0.018; single-lead sets: AUROC 0.768, SD 0.001). Considering the short amount of time needed to measure additional leads, measuring at least 3 leads—ideally more than 4 leads—is necessary for minimizing the risk of failing to detect acute myocardial infarction occurring in a certain spatial location or direction. Conclusions By developing an AI model for detecting acute myocardial infarction with asynchronous ECG lead sets, we demonstrated the feasibility of multiple lead-based AI-enabled ECG algorithms on smartwatches for automated diagnosis of cardiac disorders. We also demonstrated the necessity of measuring at least 3 leads for accurate detection. Our results can be used as reference for the development of other AI models using sequentially measured asynchronous ECG leads via smartwatches for detecting various cardiac disorders.

Digitizing ECG image: new fully automated method and open-source software code

Background: We aimed to develop and validate an automated, open-source code ECG-digitizing tool and assess agreements of ECG measurements across three types of median beats, comprised of digitally recorded, simultaneous and asynchronous ECG leads, and digitized asynchronous ECG leads. Methods: We used the data of clinical studies participants (n=230; mean age 30 ± 15 y; 25% female; 52% had the cardiovascular disease) with available both digitally recorded and printed on paper and then scanned ECGs, split into development (n=150) and validation (n=80) datasets. The agreement between ECG and VCG measurements on the digitally recorded time-coherent median beat, representative asynchronous digitized, and digitally recorded beats was assessed by Bland-Altman analysis. Results: Agreement between digitally recorded and digitized representative beat was high [area spatial ventricular gradient (SVG) elevation bias 2.5(95% limits of agreement [LOA] -7.9-13.0) degrees; precision 96.8%; inter-class correlation [ICC] 0.988; Lin concordance coefficient ρc 0.97(95% confidence interval [CI] 0.95-0.98)]. Agreement between digitally recorded asynchronous and time-coherent median beats was moderate for area-based VCG metrics (spatial QRS-T angle bias 1.4(95%LOA -33.2-30.3) degrees; precision 94.8%; ICC 0.95; Lin concordance coefficient ρc 0.90(95%CI 0.82-0.95)], but poor for peak-based VCG metrics of global electrical heterogeneity. Conclusions: We developed and validated an open-source software tool for paper-ECG digitization. Asynchronous ECG leads are the primary source of disagreement in measurements on digitally recorded and digitized ECGs.

Automated Detection of Acute Myocardial Infarction Using Asynchronous Electrocardiogram Signals—Preview of Implementing Artificial Intelligence With Multichannel Electrocardiographs Obtained From Smartwatches: Retrospective Study (Preprint)

BACKGROUND When using a smartwatch to obtain electrocardiogram (ECG) signals from multiple leads, the device has to be placed on different parts of the body sequentially. The ECG signals measured from different leads are asynchronous. Artificial intelligence (AI) models for asynchronous ECG signals have barely been explored. OBJECTIVE We aimed to develop an AI model for detecting acute myocardial infarction using asynchronous ECGs and compare its performance with that of the automatic ECG interpretations provided by a commercial ECG analysis software. We sought to evaluate the feasibility of implementing multiple lead–based AI-enabled ECG algorithms on smartwatches. Moreover, we aimed to determine the optimal number of leads for sufficient diagnostic power. METHODS We extracted ECGs recorded within 24 hours from each visit to the emergency room of Ajou University Medical Center between June 1994 and January 2018 from patients aged 20 years or older. The ECGs were labeled on the basis of whether a diagnostic code corresponding to acute myocardial infarction was entered. We derived asynchronous ECG lead sets from standard 12-lead ECG reports and simulated a situation similar to the sequential recording of ECG leads via smartwatches. We constructed an AI model based on residual networks and self-attention mechanisms by randomly masking each lead channel during the training phase and then testing the model using various targeting lead sets with the remaining lead channels masked. RESULTS The performance of lead sets with 3 or more leads compared favorably with that of the automatic ECG interpretations provided by a commercial ECG analysis software, with 8.1%-13.9% gain in sensitivity when the specificity was matched. Our results indicate that multiple lead-based AI-enabled ECG algorithms can be implemented on smartwatches. Model performance generally increased as the number of leads increased (12-lead sets: area under the receiver operating characteristic curve [AUROC] 0.880; 4-lead sets: AUROC 0.858, SD 0.008; 3-lead sets: AUROC 0.845, SD 0.011; 2-lead sets: AUROC 0.813, SD 0.018; single-lead sets: AUROC 0.768, SD 0.001). Considering the short amount of time needed to measure additional leads, measuring at least 3 leads—ideally more than 4 leads—is necessary for minimizing the risk of failing to detect acute myocardial infarction occurring in a certain spatial location or direction. CONCLUSIONS By developing an AI model for detecting acute myocardial infarction with asynchronous ECG lead sets, we demonstrated the feasibility of multiple lead-based AI-enabled ECG algorithms on smartwatches for automated diagnosis of cardiac disorders. We also demonstrated the necessity of measuring at least 3 leads for accurate detection. Our results can be used as reference for the development of other AI models using sequentially measured asynchronous ECG leads via smartwatches for detecting various cardiac disorders.

The role of single-use ECG leads in reducing healthcare-associated infections

An electrocardiogram (ECG), the recording of the electrical activity in the heart, is the most commonly performed cardiac test. It is carried out in a variety of clinical settings in hospitals and primary care, and its use is standard practice among high-risk, critically ill patients, and those who have undergone cardiac surgery. ECG recording is classified into two main categories: monitoring and diagnostic. 12-lead ECGs, which require electrodes to be placed on the chest and each limb, are used for diagnostic purposes, whereas 3- or 5-lead ECGs are used for rhythm monitoring. Cross-infection can arise from reusing ECG cables, even if they have been cleaned. Surgical site infection is a particular risk in patients who have undergone coronary artery bypass grafting, because ECG wires are placed on the chest close to the incision site. Single-use ECG leads, such as the Kendall DL™ ECG cable and lead wire system, reduce the risk of cross-contamination between patients and free nursing time for patient care because they are discarded after use and do not have to be cleaned and disinfected for use with another patient.

Spatial distribution of physiologic 12-lead QRS complex

The normal physiologic range of QRS complex duration spans between 80 and 125 ms with known differences between females and males which cannot be explained by the anatomical variations of heart sizes. To investigate the reasons for the sex differences as well as for the wide range of normal values, a technology is proposed based on the singular value decomposition and on the separation of different orthogonal components of the QRS complex. This allows classification of the proportions of different components representing the 3-dimensional representation of the electrocardiographic signal as well as classification of components that go beyond the 3-dimensional representation and that correspond to the degree of intricate convolutions of the depolarisation sequence. The technology was applied to 382,019 individual 10-s ECG samples recorded in 639 healthy subjects (311 females and 328 males) aged 33.8 ± 9.4 years. The analyses showed that QRS duration was mainly influenced by the proportions of the first two orthogonal components of the QRS complex. The first component demonstrated statistically significantly larger proportion of the total QRS power (expressed by the absolute area of the complex in all independent ECG leads) in females than in males (64.2 ± 11.6% vs 59.7 ± 11.9%, p < 0.00001—measured at resting heart rate of 60 beats per minute) while the second component demonstrated larger proportion of the QRS power in males compared to females (33.1 ± 11.9% vs 29.6 ± 11.4%, p < 0.001). The analysis also showed that the components attributable to localised depolarisation sequence abnormalities were significantly larger in males compared to females (2.85 ± 1.08% vs 2.42 ± 0.87%, p < 0.00001). In addition to the demonstration of the technology, the study concludes that the detailed convolution of the depolarisation waveform is individual, and that smoother and less intricate depolarisation propagation is the mechanism likely responsible for shorter QRS duration in females.