Comparison Between Manual and (Semi-)Automated Analyses of Esophageal Diaphragm Electromyography During Endurance Cycling in Patients With COPD

FIGURE 1. A five diaphragm EMG (EMGdi) channel recording contains electrocardiogram (ECG) artifacts during resting breathing. From top to bottom the channels are as follow; Channel 1: respiratory flow (l/s; negative flow indicating the inspiratory cycle), Channel 2: ECG recording (volt), Channels 3–7: EMGdi recordings. The correct positioning of the catheter is shown when the largest EMGdi amplitudes are in the outer EMGdi channels (3 and 7), and the smallest amplitude is in the middle channel (5). The inspiratory capacity (IC) maneuver is highlighted in the red box indicated by the higher flow, which accompanies the maximal activation of the diaphragm (EMGdiMax).

FIGURE 2. Illustration of the diaphragm EMG (EMGdi) recordings toward the end of the (symptom limited) cycling test. Channel 1: respiratory flow (l/s; negative flow indicating the inspiratory cycle), Channel 2: electrocardiogram (ECG) recording (volt), Channels 3–7: RMS EMGdi recordings. For the data analysis using the manual method, the mean values of EMGdi in between QRS complexes during the inspiratory cycle were selected. The periods highlighted in light blue are the possible periods that could be chosen without ECG contamination in each inspiratory cycle. The average of EMGdi of five consecutive breaths was used as a representative value of EMGdi of that minute. Vertical cursors 1, 2, 3, and 4 were used as a tool during manual analysis to retrieve mean values of EMGdi in the selected period.

FIGURE 3. Illustration of the diaphragm EMG (EMGdi) recordings toward the end of the (symptom limited) cycling test (period comparable to Figure 2). For the data analysis using the semi-automated method. Channel 1: respiratory flow (l/s; negative flow indicating the inspiratory cycle), Channel 2: ECG recording (volt); The absolute ECG values (volt) from this re-imported data after having been processed in LABVIEW using the semi-automatic algorithm were transformed into an abstract unit. Therefore, the signal appeared to be distorted and cannot be compared directly to the pre-processed ECG signal, Channels 3–7: the processed EMGdi data from our customized algorithm without ECG contamination in the diaphragmatic EMG signal (EMGdi). Channel 8: br.#, beginning of the inspiratory cycle and Channel 9: T exp., beginning of the expiratory cycle indicates the inspiratory and expiratory cycle of each breath which was marked automatically by the program. The EMGdi during a full inspiratory cycle can be selected (highlighted in light blue). The average of the mean EMGdi from every inspiratory cycle in each minute was analyzed. Vertical cursors 1 and 2 indicate a longer period that the value of EMGdi…

FIGURE 4. The correlation between EMGdi/EMGdiMax% calculated from the manual method by two assessors during the pre-measurement CWR cycling test (A); the semi-automated method and the manual method by assessor 1 (B); and the semi-automated method and the manual method by assessor 2 (C). Line of identity, linear regression coefficients, intra-class correlation coefficients (ICC), and significances are presented in each figure. Each of the data points represents the activation of diaphragm EMG (EMGdi/EMGdiMax%) of each patient in every minute during the pre-measurement CWR cycling test.

FIGURE 5. The results from the manual method by two assessors and the semi-automated method are compared in seven patients in each minute during the pre-measurement CWR cycling test. Bland-Altman plots of EMGdi/EMGdiMax% calculated from the manual method by two assessors (A); the semi-automated method and the manual method by assessor 1 (B); the semi-automated method and the manual method by assessor 2 (C).

FIGURE 6. Average diaphragm activation (EMGdi/EMGdiMax%) of seven patients during CWR cycling test at pre-measurement calculated from the manual method by two assessors (A); the semi-automated method and the manual method by assessor 1 (B); the semi-automated method and the manual method by assessor 2 (C).

FIGURE 7. The correlation between the pre/post difference in four participants EMGdi/EMGdiMax% calculated from the manual method by two assessors (A); the semi-automated method and the manual method by assessor 1 (B); the semi-automated method and the manual method by assessor 2 (C). Line of identity, linear regression coefficients and intra-class correlation coefficients (ICC) and significances are presented in each figure. Each of the data points represents the pre/post differences of diaphragm EMG activation (EMGdi/EMGdiMax%) pre and post the intervention of each patient in every minute during CWR cycling test.

FIGURE 8. The results from the manual method by two assessors and the semi-automated method are compared in four patients in each minute during CWR cycling test pre and post the intervention. Bland-Altman plots of the pre/post differences in EMGdi/EMGdiMax% calculated from the manual method by two assessors (A); the semi-automated method and the manual method by assessor 1 (B); the semi-automated method and the manual method by assessor 2 (C).

FIGURE 9. Average diaphragm activation (EMGdi/EMGdiMax%) of four patients during CWR cycling test pre and post the intervention calculated from the manual method by two assessors (A); the semi-automated method and the manual method by assessor 1 (B); the semi-automated method and the manual method by assessor 2 (C).

FIGURE 10. Example of EMGdi recordings during cycling exercise toward the end of the (symptom limited) cycling test at the pre- (A) and post- (B) measurement for the analysis using the manual method. Channel 1: respiratory flow (l/s; negative flow indicating the inspiratory cycle), Channel 2: ECG recording (volt), Channels 3–7: diaphragm EMG (RMS EMGdi) recordings. The periods highlighted in light blue are the possible periods that could be chosen without ECG contamination in an inspiratory cycle. Numbers in cursors regions boxes show the mean EMGdi values between two vertical cursors. The numbers in red indicate that the differences in the mean values between cursors 1–2 are closer to the values between cursor 3–4 at post-measurement.

FIGURE 11. Example of EMGdi recordings during a cycling exercise toward the end of the (symptom limited) cycling test for the analysis using the manual method. Channel 1: respiratory flow (l/s; negative flow indicating the inspiratory cycle), Channel 2: ECG recording (volt), Channels 3–7: Diaphragm EMG (RMS EMGdi) recordings. The IC maneuver is highlighted in the red circle indicated by the higher flow, which accompanies the maximal activation of the EMGdi. The periods highlighted in light blue are the periods that could be chosen without ECG contamination in an inspiratory cycle during IC maneuver. With the short inspiratory time during the IC maneuver, it left only one available (light blue) period that EMGdi could be retrieved.

FIGURE 12. Illustration of the EMGdi recordings during cycling exercise toward the end of the (symptom limited) cycling test when the inspiratory time (Ti) is shorter. Channel 1: respiratory flow (l/s; negative flow indicating the inspiratory cycle), Channel 2: ECG recording (volt), Channel 3–7: diaphragm EMG (RMS EMGdi) recordings. Several QRS complexes appear during the inspiratory cycle which is already a short period. The QRS complexes take up a large part during the inspiratory cycle resulting in less “clean” EMGdi to be selected. The red circle shows one inspiratory cycle when the QRS complex appears precisely at the peak of the amplitude of the EMGdi. The EMGdi buried under this QRS complex cannot be retrieved. Therefore, the assessor must choose the part outside of the QRS complex, which results in lower mean EMGdi value, thus underestimates the diaphragm activation. The blue highlight shows a small period of peak EMGdi. The mean EMGdi value of this period will be higher (average of high values in the short time) therefore the EMGdi being overestimated.