ecg patterns  (POWERLAB INC)

Journal: Communications Medicine

Article Title: Artificial intelligence-enabled electrocardiography contributes to hyperthyroidism detection and outcome prediction

doi: 10.1038/s43856-024-00472-4

Figure Lengend Snippet: a Components of traditional machine learning models for detecting hyperthyroidism (HT). We trained three xgboost models to predict HT using patient characteristics, ECG features, and a combination of both. The sky-blue bars represent patient characteristics, while the reddish-purple bars represent ECG features. b The area under ROC curve (AUC) of all available data on HT. This includes the DLM using ECG waveform data only and the DLM combined with patient characteristics. The sky-blue and reddish-purple bars represent the results of predictions using individual patient characteristics and ECG features, respectively. The vermillion bars represent predictions integrating features from xgboost, and the bluish-green bars represent predictions, including those from DLM. The error bars are the 95% confidence intervals (CI) of each AUC.

Article Snippet: The 35 ECG patterns were derived from statements generated by the Philips system® for each ECG.

Techniques:

Journal: Communications Medicine

Article Title: Artificial intelligence-enabled electrocardiography contributes to hyperthyroidism detection and outcome prediction

doi: 10.1038/s43856-024-00472-4

Figure Lengend Snippet: a Distribution of ECG morphologies in overt HT, subclinical HT, and non-HT groups stratified by AI-ECG. This analysis presents the differences in ECG morphologies among different groups, with each group further divided into AI-ECG(+) [representing predicted probabilities greater than the operational cutoff] and AI-ECG(−) [representing predicted probabilities less than the operational cutoff]. For continuous variables, we use boxplots to illustrate their distributions, adjusting for hospitals using linear regression. For categorical variables, we use barplots to depict proportions, adjusting for hospitals using logistic regression. Vermillion, reddish-purple, and bluish-green describe the overt HT, subclinical HT, and non-HT groups, respectively. Blue and orange represent AI-ECG(+) and AI-ECG(−). b Risk analysis of selected ECG morphologies on adverse outcomes. This analysis was conducted using the Cox proportional hazard model and combines results from all hospitals. Hazard ratios were adjusted for hospital, sex, and age. The square and error bar represent the hazard ratios and corresponding 95% confidence intervals (CI). Vermillion, black, and sky blue bars denote significantly positive, non-significant, and negative associations, respectively, with the corresponding outcomes. In this analysis, the standard deviations (SD) of heart rate, PR interval, and QRS duration were 19.5, 31.8, and 17.4, respectively.

Article Snippet: The 35 ECG patterns were derived from statements generated by the Philips system® for each ECG.

Techniques:

Journal: Communications Medicine

Article Title: Artificial intelligence-enabled electrocardiography contributes to hyperthyroidism detection and outcome prediction

doi: 10.1038/s43856-024-00472-4

Figure Lengend Snippet: Patients with a prior history of HF were excluded for the analysis of new-onset HF. Vermillion dashed line, reddish-purple dotted line, and bluish-green solid line represent the overt HT, subclinical HT, and non-HT groups, respectively. We have also highlighted the mean age for each group, as the overt HT group is relatively younger than the other groups. This age difference results in a notably higher sex and age-adjusted hazard ratio (HR), especially for new-onset HF. The table displays the at-risk population and cumulative risk for the specified time intervals, categorized by AI-ECG positive and negative.

Article Snippet: The 35 ECG patterns were derived from statements generated by the Philips system® for each ECG.

Techniques:

Journal: Communications Medicine

Article Title: Artificial intelligence-enabled electrocardiography contributes to hyperthyroidism detection and outcome prediction

doi: 10.1038/s43856-024-00472-4

Figure Lengend Snippet: a The Kaplan–Meier curve analysis stratified by AI-ECG prediction. For the new-onset HF analysis, we excluded patients with a prior history of HF. Yellow dashed line and blue solid line represent AI-ECG (+) [indicating a predicted probability greater than the operational cutoff] and AI-ECG (−) [indicating a predicted probability less than the operational cutoff], respectively. The hazard ratio (HR) presented here has been adjusted for sex and age using a Cox proportional hazards model. The table displays the at-risk population and cumulative risk for the specified time intervals in AI-ECG positive and negative patients. b Forest plot illustrating the risk of AI-ECG (+) compared to AI-ECG (−) stratified by hyperthyroidism (HT) and non-HT. The HT group includes overt and subclinical HT (Note: Some cases do not belong to either group due to a lack of free T4 results). We provide the event count and total population for each subgroup. The HR presented here was also adjusted for sex and age using a Cox proportional hazards model. In the figure, the black square represents the point estimate of the HR, while the error bars indicate the 95% confidence intervals (CI).

Article Snippet: The 35 ECG patterns were derived from statements generated by the Philips system® for each ECG.

Techniques:

Journal: Communications Medicine

Article Title: Artificial intelligence-enabled electrocardiography contributes to hyperthyroidism detection and outcome prediction

doi: 10.1038/s43856-024-00472-4

Figure Lengend Snippet: The performance is presented in bar charts and error bar, which represent the area under ROC curve (AUC) and 95% confidence intervals (CI). The analyses were stratified by data source (orange, yellow, and blue for emergency department [ED], inpatient department [IPD], and outpatient department [OPD]), sex (vermillion and sky blue for female and male), age (shades of reddish-purple from dark to light representing younger to older age), and HT information (1 and 2 representing without HT/ATD history and ECG and TSH within 24 h, in black). The black bars on the right side represent those meeting <60 y/o and conditions 1 and 2. The isolated test set was excluded from this analysis due to its small sample size. We presented the performance in internal test set ( a ) and, community test set ( b ), respectively.

Article Snippet: The 35 ECG patterns were derived from statements generated by the Philips system® for each ECG.

Techniques: Isolation

Journal: Psychopharmacology

Article Title: Myocardial infarction sensitizes medial prefrontal cortex to inhibitory effect of locus coeruleus stimulation in rats

doi: 10.1007/s00213-016-4305-y

Figure Lengend Snippet: Example of ECG waveforms collected during LC stimulation in a rat subjected to CAL (a). Three sequential ECG waveforms are shown in each row. Sequential wave forms are shown before start of stimulation (Pre-stim), during stimulation and after stimulation (Post-stim). Note that arrhythmias unique from the prestimulus baseline rhythms develop midway through stimulation and did not revert immediately upon cessation of LC stimulation. bVertical plots indicating percent bigeminy, preventricular contraction (PVC), and ventricular tachycardic (VT) rhythms detected during real or mock LC stimulation in rats subjected to CAL and sham ligation. Symbols represent percent of total waveforms (beats) detected during real or mock stimulation that were of the specified waveform subtype and distinct from any waveform detected during the prestimulus time period for each individual animal. Horizontal bar indicates group mean, and vertical bar is SEM; **P < 0.01

Article Snippet: The ECG pattern recognition software ECG-Auto (EMKA Technologies, Falls Church, VA) was used to detect arrhythmias.

Techniques: Ligation