A Deep-Learning Algorithm (ECG12Net) for Detecting Hypokalemia and Hyperkalemia by Electrocardiography: Algorithm Development

Lin, Chin; Fang, Wen; Hsu, Chia-Jung; Chen, Sy‐Jou; Huang, Kathie P.; Wang, Lin; Tsai, Chien‐Sung; Kuo, Chih Chun; Chau, Tom; Yang, Stephen J. H.; Lin, Shih Hua

doi:10.2196/15931

“…Other recent studies developed deep-learning models and tested them on large datasets to screen for hyperkalemia in patients with CKD, reporting a more robust recognition of severe hyperkalemia, thus potentially reducing the risk of sudden cardiac death in those patients 50 , 51 . However, in our study we aim, not just to detect hyperkalemia, but also in continuous quantification.…”

Section: Discussionmentioning

confidence: 99%

Monitoring blood potassium concentration in hemodialysis patients by quantifying T-wave morphology dynamics

Palmieri

¹

,

Gomis

²

,

Ferreira³

et al. 2021

Sci Rep

View full text Add to dashboard Cite

We investigated the ability of time-warping-based ECG-derived markers of T-wave morphology changes in time ($$d_{w}$$ d w ) and amplitude ($$d_a$$ d a ), as well as their non-linear components ($${d_w^{{\mathrm{NL}}}}$$ d w NL and $${d_a^{\mathrm{NL}}}$$ d a NL ), and the heart rate corrected counterpart ($$d_{w,c}$$ d w , c ), to monitor potassium concentration ($$[K^{+}]$$ [ K + ] ) changes ($$\Delta [K^+]$$ Δ [ K + ] ) in end-stage renal disease (ESRD) patients undergoing hemodialysis (HD). We compared the performance of the proposed time-warping markers, together with other previously proposed $$[K^{+}]$$ [ K + ] markers, such as T-wave width ($$T_w$$ T w ) and T-wave slope-to-amplitude ratio ($$T_{S/A}$$ T S / A ), when computed from standard ECG leads as well as from principal component analysis (PCA)-based leads. 48-hour ECG recordings and a set of hourly-collected blood samples from 29 ESRD-HD patients were acquired. Values of $$d_w$$ d w , $$d_a$$ d a , $${d_w^{\mathrm{NL}}}$$ d w NL , $${d_a^{\mathrm{NL}}}$$ d a NL and $$d_{w,c}$$ d w , c were calculated by comparing the morphology of the mean warped T-waves (MWTWs) derived at each hour along the HD with that from a reference MWTW, measured at the end of the HD. From the same MWTWs $$T_w$$ T w and $$T_{S/A}$$ T S / A were also extracted. Similarly, $$\Delta [K^+]$$ Δ [ K + ] was calculated as the difference between the $$[K^{+}]$$ [ K + ] values at each hour and the $$[K^{+}]$$ [ K + ] reference level at the end of the HD session. We found that $$d_{w}$$ d w and $$d_{w,c}$$ d w , c showed higher correlation coefficients with $$\Delta [K^+]$$ Δ [ K + ] than $$T_{S/A}$$ T S / A —Spearman’s ($$\rho$$ ρ ) and Pearson’s (r)—and $$T_w$$ T w —Spearman’s ($$\rho$$ ρ )—in both SL and PCA approaches being the intra-patient median $$\rho \ge 0.82$$ ρ ≥ 0.82 and $$r \ge 0.87$$ r ≥ 0.87 in SL and $$\rho \ge 0.82$$ ρ ≥ 0.82 and $$r \ge 0.89$$ r ≥ 0.89 in PCA respectively. Our findings would point at $$d_{w}$$ d w and $$d_{w,c}$$ d w , c as the most suitable surrogate of $$\Delta [K^+]$$ Δ [ K + ] , suggesting that they could be potentially useful for non-invasive monitoring of ESRD-HD patients in hospital, as well as in ambulatory settings. Therefore, the tracking of T-wave morphology variations by means of time-warping analysis could improve continuous and remote $$[K^{+}]$$ [ K + ] monitoring of ESRD-HD patients and flagging risk of $$[K^{+}]$$ [ K + ] -related cardiovascular events.

show abstract

“…The application of deep learning technology in the cardiovascular eld for arrhythmias, dyskalemia, and valvular heart disease had become popularized recently. [19][20][21][27][28][29] However, no large scale study has been designed to apply deep learning technology for MI detection. Previous DLMs for MI detection by ECG were analyzed mainly from the Physikalisch-Technische Bundesanstalt (PTB) diagnostic ECG Database.…”

Section: Discussionmentioning

confidence: 99%

“…The technology details, such as the model architecture, data augmentation, and model visualization, were described previously. [21] We used the same architecture to train two new deep learning models for MI detection and location analysis of STEMI. The rst deep learning model was trained via full samples with 3 categories, including STEMI, NSTEMI, and not-MI, and the output of this model was a 3-class softmax output.…”

Section: Implementation Of the Dlmmentioning

confidence: 99%

See 1 more Smart Citation

A Deep-learning Algorithm With the Real World Validation for Detecting Acute Myocardial Infarction

Liu¹,

Lin²,

Tsai³

et al. 2020

Preprint

Self Cite

View full text Add to dashboard Cite

BackgroundThe initial detection and diagnosis of ST-segment or non-ST-segment elevation myocardial infarction (STEMI or NSTEMI) definitely rely on a 12-lead electrocardiogram (ECG). Delay or misdiagnosis is not unusual by subjective interpretation. Our aim is to develop a DLM as a diagnostic support tool to detect MI based on a 12-lead ECG and to evaluate the performance of this model.MethodsThis study included 1,051 ECGs from 737 coronary angiography (CAG)-validated STEMI patients, 697 ECGs from 287 CAG-validated NSTEMI patients, and 140,336 not-MI ECGs from 76,775 patients at emergency departments. DLM was trained and validated for the performance using 80% and 20% of the ECGs, respectively. A human-machine competition was conducted. The area under the receiver operating characteristic curve (AUC), sensitivity, and specificity were used to evaluate the performance of DLM and experts. STEMI versus not-STEMI, and MI versus not-MI were evaluated by DLM.ResultsThe AUCs of DLM for identifying STEMI and MI were 0.976 and 0.944 in the human-machine competition, respectively, which were significantly better than those of our best clinicians. In the real world setting, DLM presented with AUC of 0.995/0.916 with corresponding sensitivities of 96.9%/77.0%, and specificities of 96.2%/92.9% in the identification of STEMI and MI, respectively. Furthermore, DLM demonstrated sufficient diagnostic capacity for STEMI without the aid of troponin I (TnI) (AUC= 0.996) with corresponding sensitivity and specificity of 98.4% and 96.9%. The AUC of combined DLM and the first recorded TnI for the detection of NSTEMI were increased to 0.978 with corresponding sensitivity and specificity of 91.6% and 96.7%, which was better than that of DLM (0.877) or TnI (0.949) alone. ConclusionsDLM may serve as a diagnostic decision tool to assist intensive or emergency medical system-based networks and frontline physicians in identifying STEMI and NSTEMI in a timely and precise manner to prevent delay or misdiagnosis, and thereby to facilitate subsequent reperfusion therapy.

show abstract

“…We developed a deep learning model called ECG12Net to use 12 lead ECGs for potassium concentration prediction. The technology details, such as model architecture, data augmentation, and model visualization, were described previously [ 21 ].…”

Section: Methodsmentioning

confidence: 99%

Detecting Digoxin Toxicity by Artificial Intelligence-Assisted Electrocardiography

Chang

¹

,

Lin

²

,

Tsao

³

et al. 2021

IJERPH

Self Cite

View full text Add to dashboard Cite

Although digoxin is important in heart rate control, the utilization of digoxin is declining due to its narrow therapeutic window. Misdiagnosis or delayed diagnosis of digoxin toxicity is common due to the lack of awareness and the time-consuming laboratory work that is involved. Electrocardiography (ECG) may be able to detect potential digoxin toxicity based on characteristic presentations. Our study attempted to develop a deep learning model to detect digoxin toxicity based on ECG manifestations. This study included 61 ECGs from patients with digoxin toxicity and 177,066 ECGs from patients in the emergency room from November 2011 to February 2019. The deep learning algorithm was trained using approximately 80% of ECGs. The other 20% of ECGs were used to validate the performance of the Artificial Intelligence (AI) system and to conduct a human-machine competition. Area under the receiver operating characteristic curve (AUC), sensitivity, and specificity were used to evaluate the performance of ECG interpretation between humans and our deep learning system. The AUCs of our deep learning system for identifying digoxin toxicity were 0.912 and 0.929 in the validation cohort and the human-machine competition, respectively, which reached 84.6% of sensitivity and 94.6% of specificity. Interestingly, the deep learning system using only lead I (AUC = 0.960) was not worse than using complete 12 leads (0.912). Stratified analysis showed that our deep learning system was more applicable to patients with heart failure (HF) and without atrial fibrillation (AF) than those without HF and with AF. Our ECG-based deep learning system provides a high-accuracy, economical, rapid, and accessible way to detect digoxin toxicity, which can be applied as a promising decision supportive system for diagnosing digoxin toxicity in clinical practice.

show abstract

A Deep-Learning Algorithm (ECG12Net) for Detecting Hypokalemia and Hyperkalemia by Electrocardiography: Algorithm Development

Cited by 75 publications

References 42 publications

Monitoring blood potassium concentration in hemodialysis patients by quantifying T-wave morphology dynamics

Monitoring blood potassium concentration in hemodialysis patients by quantifying T-wave morphology dynamics

A Deep-learning Algorithm With the Real World Validation for Detecting Acute Myocardial Infarction

Detecting Digoxin Toxicity by Artificial Intelligence-Assisted Electrocardiography

Contact Info

Product

Resources

About