Diagnostic accuracy of a new detection algorithm for atrial fibrillation in cardiac telemonitoring with portable electrocardiogram devices.

BACKGROUND: Telemedical approaches targeting cardiac outpatients try to include electrocardiogram (ECG) analysis. Increasing numbers of monitored patients require automated preanalysis of the ECG to prioritize the evaluation for the clinical professional to enable an efficient intervention. METHODS: ECGs were recorded from 60 patients, both with a standard 12-lead ECG and with a new handheld ECG device having dry electrodes for direct skin contact. Recordings of the handheld device were automatically analyzed by a new algorithm. The 12-lead recordings were evaluated by a blinded cardiologist and then compared to the automated analysis of the handheld ECG. Sensitivity and specificity of the algorithm for the detection of atrial fibrillation (AF) were calculated. RESULTS: A total of 60 ECG strips having 122 ± 36 beats were registered. One hundred percent of the ECG strips were sufficient for automated heart rate count; 96.6%, for automated AF analysis; and 80%, for PQ, QRS, and QTc time measurements. AF detection had a sensitivity of 92.9% and a specificity of 90.9%. There was no difference in heart rate count between automated and manual analysis (median, 71 vs 70 beats per minute; P = .51). Automated measurements of a summary complex showed no difference for PQ time (165 vs 161 milliseconds, P = .50) but overestimated QRS (119 vs 90 milliseconds, P = .001) and QTc (489 vs 417 milliseconds, P < .001) times as compared to the 12-lead recordings analyzed manually. CONCLUSION: The new algorithm is suitable for automated preanalysis of the ECG data with regard to AF. It could be used for rapid selection of ECGs with relevant rhythm abnormalities from a large pool. Electrocardiographic data remain to be evaluated by health care professionals for exact diagnosis.

Analysis of atrial fibrillatory activity from high-resolution surface electrocardiograms: Evaluation and application of a new system.

BACKGROUND: Algorithms of signal processing allow the estimation of atrial fibrillation (AF) activity from surface electrocardiograms (ECGs). OBJECTIVE: To evaluate a new commercially available ECG system for AF analysis from surface potentials. METHODS: Patients (n=52, mean [+/- SD] age of 68+/-9.6 years) with persistent AF (mean duration 44+/-52.2 months), referred for cardiac surgery, underwent high-gain, high-resolution ECG preoperatively. After QRST cancellation, the frequency content of AF was identified by fast Fourier transformation. Epicardial potentials were registered at the right atrial appendage, the right atrium (RA), the left atrium (LA) and the left atrial appendage intraoperatively (nine patients). RESULTS: Mean (+/- SD) fibrillatory rate (MFR) in lead V(1) (393+/-40.4 fibrillations/min [fpm]) correlated significantly with V(2) (391+/-43.3 fpm, r=0.976; P<0.05) and II (379+/-41.1 fpm, r=0.878; P<0.05), and was stable within an interval of 13.6+/-3.8 min (27 patients). In the right atrial appendage, RA, LA and left atrial appendage, the relative difference in MFR was small (5.6%, 6.6%, 6.8% and 5.7%, respectively, compared with V(1)). The mean peak frequency component at 75% of the maximum power was significantly smaller in the LA than in the RA (13+/-4.2 fpm versus 22+/-7.2 fpm, respectively; P<0.01), and in patients with high (more than 390 fpm) compared with low (390 fpm or fewer) fibrillatory activity (14+/-7.6 fpm versus 22+/-13.3 fpm, respectively; P<0.05). There was a nonsignificant trend to higher fibrillatory activity with longer AF duration. Other characteristics (age, sex, LA size, ejection fraction, type of heart disease and medication) were not associated with the MFR. CONCLUSIONS: Using the CardioLink system, AF analysis from surface ECG is reliable and equivalent to epicardial measurements. By noninvasive assessment of individual electrical remodelling, this system certainly supports clinical AF research.

The diagnostic path, a useful visualisation tool in virtual microscopy.

BACKGROUND: The Virtual Microscopy based on completely digitalised histological slide. Concerning this digitalisation many new features in mircoscopy can be processed by the computer. New applications are possible or old, well known techniques of image analyses can be adapted for routine use. AIMS: A so called diagnostic path observes in the way of a professional sees through a histological virtual slide combined with the text information of the dictation process. This feature can be used for image retrieval, quality assurance or for educational purpose. MATERIALS AND METHODS: The diagnostic path implements a metadata structure of image information. It stores and processes the different images seen by a pathologist during his "slide viewing" and the obtained image sequence ("observation path"). Contemporary, the structural details of the pathology reports were analysed. The results were transferred into an XML structure. Based on this structure, a report editor and a search function were implemented. The report editor compiles the "diagnostic path", which is the connection from the image viewing sequence ("observation path") and the oral report sequence of the findings ("dictation path"). The time set ups of speech and image viewing serve for the link between the two sequences. The search tool uses the obtained diagnostic path. It allows the user to search for particular histological hallmarks in pathology reports and in the corresponding images. RESULTS: The new algorithm was tested on 50 pathology reports and 74 attached histological images. The creation of a new individual diagnostic path is automatically performed during the routine diagnostic process. The test prototype experienced an insignificant prolongation of the diagnosis procedure (oral case description and stated diagnosis by the pathologist) and a fast and reliable retrieval, especially useful for continuous education and quality control of case description and diagnostic work. DISCUSSION: The Digital Virtual Microscope has been designed to handle 1000 images per day in the daily routine work of a pathology institution. It implies the necessity of an automatic mechanism of image meta dating. The non - deterministic correlation between the oral statements (case report) and image information content guides the image meta dating. The presented software opens up new possibilities for a content oriented search in a virtual slide, and can successfully support medical education and diagnostic quality assurance.

Detection of acute myocardial ischemia during percutaneous transluminal coronary angioplasty by endocardial acceleration.

The first heart sound is generated by vibrations from the myocardium during isovolumic contraction. Peak endocardial acceleration (PEA) has been used previously to measure these vibrations in humans and correlates with myocardial contractility during inotropic interventions. It is unknown if changes in PEA can be used to characterize a reduction in contractility during ischemic episodes. This study was designed to evaluate the use of an endocardial accelerometer for the detection of acute myocardial ischemia. Thirteen patients undergoing routine percutaneous transluminal coronary angioplasty (PTCA) consented to having a single-axis, lead-based accelerometer positioned in the right ventricular apex. PEA was defined as the maximum peak-to-peak amplitude during a window 50 ms before to 200 ms following the peak R wave. Time of endocardial acceleration (TEA) was defined as the time from the peak R wave to the maximum accelerometer signal within this window. To obtain a more robust estimate of the strength of vibrations, a 100-beat template of the accelerometer signal was constructed at baseline and applied as a matched filter during ischemia. The peak magnitude of the filtered endocardial accelerometer signal (Max Filtered EA) was used as an index of signal intensity. Median baseline PEA, TEA, and Max Filtered EA were 0.91 +/- 0.35 g, 75.2 +/- 16.2 ms, and 0.40 +/- 0.20 g, respectively. PEA and Max Filtered EA significantly decreased by 7% during ischemia (0.91 to 0.85 g and 0.40 to 0.37 g, both P < 0.05, respectively). TEA did not significantly change from baseline (77.0 ms, P = ns). The results of this study suggest that acute ischemia can be detected with an endocardial accelerometer in humans.

Comparison of electrocardiogram and intrathoracic electrogram signals for detection of ischemic ST segment changes during normal sinus and ventricular paced rhythms.

INTRODUCTION: The aim of this study was to compare surface ECGs with electrograms (EGM) that are available from implanted devices for the ability to detect ischemic ST segment changes during normal sinus (NS) and ventricular paced (VP) rhythms. METHODS AND RESULTS: ECG leads I, II, and V2, right atrial ring to left pectoral patch (representing the can of the device), right ventricular ring to left pectoral patch, and right atrial ring to right ventricular ring EGM were recorded continuously during percutaneous transluminal coronary angioplasty. One balloon inflation (> or = 60 sec) was analyzed from each of 22 NS and 22 VP subjects. The parameter AST was defined as the maximum absolute ST segment deviation (from isoelectric) during the first 60 seconds of inflation, measured relative to the baseline (preinflation) ST segment deviation. For EGM, a normalized deltaST was defined as the AST divided by the ratio of QRS amplitudes of EGM to ECG. During NS, the deltaST for EGM (0.43 mV) was significantly larger than that of ECG (0.09 mV, P = 0.0001) but the normalized deltaST for EGM (0.11 mV) was comparable to that of ECG (0.09 mV, P = 0.45). During VP, the AST for EGM (1.08 mV) was significantly larger than that of ECG (0.17 mV, P = 0.0001), but the normalized AST for EGM (0.11 mV) was significantly smaller than that of ECG (0.17 mV, P = 0.02). CONCLUSION: During both NS and VP, ischemic ST segment changes were significantly larger in EGM than in ECG. Much of this difference appears to be related to larger amplitudes of EGM signals. (J