Creation of a training and test dataset with the disposition and transposition of overlaying electrocardiographic electrodes when recording electrocardiograms-12

Full Text

BACKGROUND

Electrocardiography (ECG) is one of the simplest, most widespread, inexpensive, and informative techniques in cardiology; however, its diagnostic value is dramatically reduced if it is not performed correctly. The most common change in the recording technique is electrode displacement (dislocation). A disposition (a deliberate and necessary change in electrode placement) should be distinguished from a transposition (an erroneous change in electrode placement compared with the standard overlay pattern).

Several studies have attempted to systematize errors and deviations in electrode placement; however, all of them concerned the most frequent variants, such as relocation of red and yellow, yellow and green, and precordial electrodes above or below the standard pattern. [1-7] To avoid errors in the interpretation of ECG recordings, algorithms capable of detecting such errors are developed. Machine algorithms stably recognize right–left (RL) transposition.

To assess the quality and determine the criteria for the possibility of using such algorithms, ECG datasets should be created for training and testing both algorithms for automatic ECG analysis and by healthcare specialists, i.e., doctors and nursing staff.

This study aimed to create an ECG dataset with different electrode transpositions and dispositions during ECG recording for training and testing machine-learning systems.

MATERIALS AND METHODS

Inclusion criteria

Patients aged 18–75 years (27 men and 22 women) were included in the study. All patients gave voluntary informed consent for ECG recording.

Settings

The study was performed in the Moscow State Budgetary Institution City Clinical Hospital No. 67 named after L.A. Vorokhobov of Moscow Healthcare Department.

Description of treatment

The patients were divided into the following six arms:

1) Patients with normal ECG (recordings starting with 101, 102, 103, 104, 105, 106, 107, 108, 109, and 110).

2) Patients with ECG signs of left ventricular hypertrophy or complete left bundle branch block (recordings starting with 201, 202, 203, 204, 205, 206, 207, 208, and 209).

3) Patients with ECG signs of right bundle branch block (recordings starting with 301, 302, 303, 304, 305, and 306).

4) Patients with recorded ST (−) depression (recordings starting with 401, 402, 403, 404, 405, 406, 407, 408, and 409).

5) Patients with recorded ST(+) elevation, including the phenomenon of early ventricular repolarization (recordings starting with 501, 502, 503, 504, 505, 506, and 507).

6) Patients with any nonsinus rhythm that are preferably characterized by negative P waves in I, II, V1, and V2 leads (recordings starting with 601, 602, 603, 604, 605, 606, 607, and 608).

Cardiograms were recorded using EASY ECG Modular System for Recording and Remote Transmitting of ECGs according to TU 9441-001-42447560-2012 with Accessories (ATES MEDICA SOFT LTD, Russia, Registration Certificate No. RZN 2018/7062).

For ECG recording, operating silver/silver chloride electrodes were used with 26 × 47-mm plates (for limb ones) and 22-mm cups (for precordial ones).

Digitized cardiac signals were recorded in the European data format. [8] Recordings were made in 12 common leads with a sample rate of 500 Hz, the recording lasted for 10 s, and the signal magnitude was 10 mV. No signal filtering was performed, and the bandwidth was from 0.05 to 150 Hz.

Ethical review

The study was approved by the independent ethical committee of the Moscow Regional Branch of the Russian Society of Roentgenologists and Radiologists (RSRR MRB IEC).

RESULTS

Each patient underwent sequential ECG recording during one visit with correct electrode placement and different variants of dis- and transpositions. The coding and description of overlay patterns are provided in Table 1.

 

Table 1. Electrode placement variants and their coding

Russian wording	English equivalent	Part of the ECG filename	Comment	Incorrectly displayed leads
Стандартное расположение электродов	Standard electrodes’ arrangement	dis_st	Correctly applied electrodes	All leads are correct
Грудные электроды: на 2 межреберья выше стандартной схемы	Precordial electrodes’ misplacement: up to 2 intercostal spaces above the standard arrangement	dis_u2	Occasionally required for clinical indications, but in general, it is the most common electrode misplacement	Correct: I, II, III, aVR, aVL, and aVF. Dislocated: V1, V2, V3, V4, V5, and V6
Грудные электроды: на 2 межреберья ниже стандартной схемы	Precordial electrodes’ misplacement: below to 2 intercostal spaces down the standard arrangement	dis_d2	Occasionally required for clinical indications, but in general, it is electrode misplacement	Correct: I, II, III, aVR, aVL, and aVF. Dislocated: V1, V2, V3, V4, V5, and V6
Перепутаны красный (R) и жёлтый (L) электроды	Reversal of the two arm electrodes	trns_ry	Misplacement	Incorrect: I, II, III, aVR, aVL, and aVF. Correct: V1, V2, V3, V4, V5, and V6
Перепутаны жёлтый (L) и зелёный (F) электроды	Reversal of the left arm and left leg electrodes	trns_yg	Misplacement	Incorrect: I, II, III, aVR, aVL, and aVF. Correct: V1, V2, V3, V4, V5, and V6
Все грудные электроды расположены на прямой линии от C1–C6	All chest electrodes are placed at the same line	trns_ln	Misplacement	Correct: I, II, III, aVR, aVL, aVF, and V1. Incorrect: V2, V3, V4, V5, and V6
Перепутаны C1–C2	Reversal of the C1 and C2 electrodes	trns_12c	Misplacement	Correct: I, II, III, aVR, aVL, aVF, V3, V4, V5, and V6. Incorrect: V1 and V2
Перепутаны C5–C6	Reversal of the C5 and C6 electrodes	trns-56c	Misplacement	Correct: I, II, III, aVR, aVL, aVF, V1, V2, V3, and V4. Incorrect: V5 and V6
Электроды C4, C5, C6 расположены очень близко друг к другу (соприкасаются)	The C4, C5, C6 electrodes are placed too close to each other (cheek to cheek)	trns_cls	Misplacement	Correct: I, II, III, aVR, aVL, aVF, V1, V2, and V3. Incorrect: V4, V5, and V6
Электроды C4, C5, C6 расположены по межреберью (не по горизонтальной линии, а искривляются вверх)	The C4, C5, C6 electrodes are misplaced along the intercostal space (not at a horizontal line, but curved upwards)	trns_crv	Misplacement	Correct: I, II, III, aVR, aVL, aVF, V1, V2, and V3. Incorrect: V4, V5, and V6

 

In total, 486 ECGs were recorded in 49 patients. The structure of the recorded ECGs is shown in Table 2.

 

Table 2. Allocation of cardiograms recorded in patients

Placement type Subgroup of ECG changes	Standard ECG placement	Dislocation of precordial electrodes	Transposition of limb electrodes	Transposition of precordial electrodes	Total
Normal ECG	10	20	20	49	99
ECG with LVH and CLBBB	9	18	18	45	90
ECG with RBBB	6	12	12	30	60
ECG with ST(−) depression	9	18	18	43	88
ECG with ST(+) elevation	7	14	14	35	70
ECG with any nonsinus rhythm	8	15	16	40	79
Total	49	97	98	242	486

Notes. ECG, electrocardiogram; CLBBB, complete left bundle branch block; LVH, left ventricular hypertrophy; RBBB, right bundle branch block.

 

After recording, all results were anonymized and pseudonymized¹, placed, and annotated on the ECG.RU platform using a unified thesaurus. [9] Each ECG file was named according to the following principles: the first three digits referred to the patient subgroup number and serial number, and subsequent letters referred to the coding of electrode placement.

For example, the file name “101_dis_st” meant that this ECG was recorded in a patient with an unchanged ECG, and the electrodes were placed correctly. The file name “203_trns_crv” contains the ECG of a patient with left ventricular hypertrophy, and the C4, C5, and C6 electrodes are placed along the intercostal line (not along the horizontal line, but curved upward). The file name “602_trns_yg” contains the ECG of a patient with nonsinus rhythm, and the yellow and green electrodes are reversed. Examples of recorded ECGs are shown in pairs in Figures 1–3 (overlay pattern and corresponding ECG).

 

Fig. 1. Misplacement of L and F electrodes.

 

Fig. 2. Misplacement of C4–C6 electrodes.

 

Fig. 3. Dislocation of C4–C6 electrodes.

 

DISCUSSION

The obtained results indicate a significant variability of ECG patterns depending on the initial changes and, probably, the patient’s body type and state of mammary glands. Despite sufficient studies on the variability of normal ECG changes, [6–8] the variability of changes in the initial abnormal ECG is yet to be examined, which we will do in our next paper.

In the visual evaluation of the recorded ECGs, transposition was determined based on the relocation of the arm leads (red and yellow—RL) and precordial C1–C2 was not complicated. The placement of precordial cheek-to-cheek electrodes, in disposition with the transfer of the precordial leads higher or lower by two intercostal spaces than the Wilson scheme (upper2 and lower2), was less often reliably determined. Transpositions of yellow and green limb electrodes (LF), changes in the position of precordial electrodes, when they are in line, curved, or C5–C6 electrodes are reversed, are difficult to determine even when directly comparing two cardiograms, with correct placement and transposition of electrodes. This probably depends on baseline ECG changes, body type (normosthenic, hypersthenic, or asthenic), breast size, or presence of an implant. Certainly, issues such as the consideration of the built, presence of a transplant, possible detailing of the initial ECG changes according to the reduced thesaurus, and comparison of the efficiency of recognition by a specialist and automatic analysis algorithms, require further evaluations.

Despite the known disadvantages, considering the high importance of the issue and difficulty of obtaining qualitatively recorded data and annotated ECGs, the obtained set of ECGs was published under an open license and used for both verification of algorithms of automatic ECG analysis and training of doctors and nursing staff.

Study limitations

On patient examination, their height, weight, and built (somatotypes) were not always considered. Taking these factors into account will probably increase the accuracy of determining the type of electrode dislocation.

CONCLUSION

In this study, an ECG dataset with different electrode dislocation variants was obtained. The dataset consists of ECG series recorded in each patient with different electrode overlap variants. The set includes not only normal ECGs but also different variants of abnormal ECGs.

ADDITIONAL INFORMATION

Funding source. This article was not supported by any external sources of funding.

Competing interests. The authors declare that they have no competing interests.

Authors’ contribution. All authors made a substantial contribution to the conception of the work, acquisition, analysis, interpretation of data for the work, drafting and revising the work, final approval of the version to be published and agree to be accountable for all aspects of the work. D.V. Drozdov ― concept development, organization of annotation process, thesaurus development, author supervision; T.M. Gazashvili ― organization of ECG registration, depersonalization and transfer of records; D.V. Shutov ― concept development, organization of depersonalization, transfer, annotation, author supervision; A.S. Skoda ― general management, organization of the work of the center for instrumental diagnostics.

Acknowledgments. The authors express their gratitude to Medical Computer Systems LLC, personally to the director of the company Prilutsky Dmitry Anatolyevich and all employees of the platform ECG.ru. Meshcherina Tatiana Evgenievna and Shalmonova Elena Gennadievna for highly qualified ECG registration.

¹ GOST R 55036-2012/ISO/TS 25237:2008. P85 Group. National Standard of the Russian Federation. Health Informatics. Pseudonymization. Access: https://docs.cntd.ru/document/1200100339.

About the authors

Tamara M. Gazashvili

City Clinical Hospital No. 67 named after L.A. Vorokhobov

Email: Tamaradoc24@gmail.com
ORCID iD: 0000-0002-5875-9699
SPIN-code: 4208-2303
Russian Federation, Moscow

Dmitry V. Drozdov

National Medical Research Centre of Cardiology Named After Academician E.I. Chazov

Email: cardioexp@gmail.com
ORCID iD: 0000-0001-7374-3604
SPIN-code: 2279-9657

MD, Cand. Sci. (Med)

Russian Federation, Moscow

Dmitry V. Shutov

Research and Practical Clinical Center for Diagnostics and Telemedicine Technologies

Email: ShutovDV@zdrav.mos.ru
ORCID iD: 0000-0003-1836-3689
SPIN-code: 9381-2456

MD, Dr. Sci. (Med)

Russian Federation, Moscow

Andrey S. Shkoda

City Clinical Hospital No. 67 named after L.A. Vorokhobov

Author for correspondence.
Email: a.shkoda@67gkb.ru
ORCID iD: 0000-0002-9783-1796
SPIN-code: 4520-2141

MD, Dr. Sci. (Med)

Russian Federation, Moscow

References

Supplementary files