Noninvasive electromyometrial imaging of human uterine maturation during term labor

Development and implementation of the human EMMI system

The human EMMI system incorporates subject-specific body-uterus geometry and multiple channel electrical measurements (up to 192 electrodes) from the body surface to reconstruct complete electrical activities in sequential frames across the 3D uterus.

In this study, human EMMI is performed on term labor subjects in three steps. First, a subject at ~37 weeks’ gestation undergoes an MRI scan while wearing up to 24 MRI patches containing up to 192 MRI-compatible fiducial markers around the body surface (Fig. 1a). Second, in the first stage of labor, when cervical dilation is at least 3 cm, and regular contractions are observed from the TOCO monitor, customized pin-type electrode patches are applied to the same locations on the body surface as the MRI fiducial markers. Because clinical devices such as a TOCO monitor and a fetal monitor must be applied to the abdomen to guide clinical decisions, the locations of the electrode patches were adjusted. To locate the electrode positions on an individual basis, an optical 3D scanner is used to record the actual electrode positions (Fig. 1b). Third, the subject undergoes body surface electrical recording (Fig. 1c). Each recording lasts ~15 minutes, and four recordings (up to one-hour total) are conducted for each subject in this study. The temporal sampling rate is 2048 Hz.

a An MRI scan is performed while the subject is wearing up to 24 MRI patches containing up to 192 markers. b An optical 3D scan of the body surface is performed while the subject is wearing electrode patches in the corresponding positions as the MRI patches. c Body surface electromyograms are simultaneously recorded from up to 192 pin-type unipolar electrodes assembled as patches. EMMI generates d, a body-uterus geometry from MR images with electrode locations on the body surface and, e a body surface potential map by rendering the electromyograms at each electrode at an instant in time on the body surface. EMMI combines the two data sets to reconstruct f uterine surface potential maps (electrical activity across the uterus at a single time point). With the potential maps, we can generate the electrograms g electrical waveforms over time at each uterine site, and then derive h, the uterine region, and chronological sequence of electrical activation visualized as isochrone maps in 3D. 3D three-dimensional, MRI magnetic resonance imaging, EMMI electromyometrial imaging.

Raw data of MR images, optical 3D scanning, and body surface electromyograms (EMG) are preprocessed to generate a subject-specific body-uterus geometry (Fig. 1d) and body surface potential maps over the body surface (Fig. 1d). The body-uterus geometry includes the coordinates of the body surface electrode locations (blue dots in Fig. 1d) and the discretized uterine surface site locations (See body-uterus geometry construction in Methods). Filtering and artifact removal is applied to the raw EMG recording to improve the signal-to-noise ratio (See EMG signal preprocessing in Methods).

The Method of fundamental solutions¹⁶ was employed to solve the three-dimensional Cauchy problem to generate uterine potential maps (electrical activity distribution on the uterine surface as a function of time every 10 milliseconds, Fig. 1f). These potential maps are essentially a 4D time-series data set: electrograms (Fig. 1g, D time series data over the entire recording period) at multiple sites on the uterine surface in 3D. During a contraction, we determine the uterine electrical activation times by measuring the start times of uterine electrogram burst (UEB) at each uterine surface site, which will be used to form the isochrone map (Fig. 1h). In the isochrone map, warm colors denote regions of the uterus that are electrically activated earlier during a contraction, cool colors denote regions that are activated later, and gray regions that are inactivated. As shown in Fig. 1g and Fig. 1h, the EMMI uterine surface electrograms reflect the local electrical activities. For example, in Fig. 1g, the electrogram on the right has a UEB (where UEBs can be detected above the baseline at an SNR > 5 Decibels) that maps to a site (marked with a plus sign in Fig. 1h) in the isochrone map near the region of the uterus that is first activated. In contrast, the electrogram on the left does not show a UEB and maps to a site (marked with an asterisk) in the isochrone map that showed no electrical activity during the contraction. EMMI system (MRI scanner, optical 3D scanner, and electrode patches) is described in detail in the Methods section.

Noninvasive imaging of human labor contractions

A body surface EMG measured from the body surface of one representative subject (Subject #2) is shown in Fig. 2a. One recording segment from ~18th to 41st second was magnified. Figure 2b through Fig. 2e show sequential potential maps on the body surface and uterine surface in the anterior view in the time windows labeled as b, c, d, and e in Fig. 2a, respectively. At each indicated time window, the body surface potential maps (body in Fig. 2b) were generated from the multichannel body surface electrode measurements. In comparison, the uterine surface potential maps (uterus in Fig. 2b) were reconstructed by EMMI. Unlike conventional EMG techniques in that the electrical activities are measured from the body surface, EMMI incorporates the subject-specific body-uterus geometry to generate sequential potential maps across the entire 3D uterine surface with high temporal resolution. The EMMI uterine surface potential maps enable noninvasive characterization of the electrical activities distributed over the entire uterine surface of a subject, and allow the detection of local electrical activities in the myometrium with high spatial resolution. Currently, the uterine surfaces consist of 320 vertices evenly, and the spatial resolution is ~2.5 cm.

a A representative body surface EMG was measured at the location labeled as star (*) in Subject #2 experiencing a uterine contraction in the latent phase of labor. One segment from ~18th second to 41st second was magnified. b–e Sequential potential maps on the body surface and uterine surface in anterior view at the indicated times. Each row corresponds to the time window labeled as b–e in a, respectively. EMG electromyogram. Source data are provided as a Source Data file.

The 4D spatial-temporal uterine surface potential map imaged by EMMI (in Figs. 2b through 2e) reveals the evolution of phase and magnitude inside a UEB. The uterine potential map can be reorganized into multichannel uterine surface electrograms based on the spatial locations on the 3D uterine surface. Each EMMI uterine surface electrogram reflects the local electrical activities of one uterine site. Based on the EMMI uterine electrograms during a uterine contraction, we define the electrical activation on the uterine surface by the initiation of a UEB. The term “uterine electrical activation time” or just “activation time” is used here to refer to the initiation time of the UEB. For the same representative subject shown in Fig. 2, simultaneous TOCO signal and five representative uterine surface electrograms (A–E in Fig. 3b) from five uterine surface sites (Fig. 3d) during two consecutive uterine contractions were shown in Fig. 3a, b, respectively. For each EMMI uterine surface electrogram, UEBs were detected (Supplementary Fig. 2), labeled by the upward red step lines, and overlaid on the electrograms. The rising edge of the red step line indicates the activation time during the contraction (green arrows). During the first uterine contraction, uterine surface electrograms from sites A through D demonstrated clear UEBs, suggesting that uterine sites A to D were electrically activated. In comparison, no UEB was detected in the uterine electrogram from site E, indicating that the myometrium around site e was inactive. Thus the entire myometrium was not electrically active and contributed to the uterine contractions. Inspecting all uterine surface sites, the earliest activation and the latest activation times can be detected and are marked by the dashed black vertical lines in Fig. 3a, b. EMMI uterine surface electrograms also suggest that the activation sequence among different uterine sites can change from one contraction to the next. For example, site B activated earlier than sites C and D in the first contraction, while site B activated after sites C and D in the second one (Fig. 3a).

a For a contraction in Subject #2 that the clinical TOCO monitor confirmed, b the electrical activations are defined as the initiation of UEB in the multichannel electrograms at about 320 uterine sites covering the entire uterine surface. In the 5 representative uterine electrograms from the indicated sites marked A through E in d, red step lines denote the UEB, green arrows denote the electrical activations, and the dashed black lines denote the earliest and latest electrical activations. c The entire activation process is visualized by the generation process of the uterine activation isochrone map and the activation curve. The former shows the activation location and time across the 3D uterine surface; the latter shows how the activation ratio increases over time. It occurs during the first part of the TOCO signal of contraction. d The complete isochrone map reflects the electrical activity of the myometrium in time and space during the contraction, where warm colors denote uterine regions that are activated early, cool colors denote the regions that are activated late, and gray denotes the regions that are never inactivated. e In the early activation map, inactive regions are in gray, and activated regions are divided into two parts: early activation (red) the 33% of areas that are activated first in time, and late activation (blue) the remaining 67%. The fundal boundary is labeled as a dashed white curve. f The activation curve reflects the temporal progression of the electrical activation during the contraction. EMMI indices (MAR, ACS, and FAR) quantify the electrical properties of the myometrium. TOCO tocodynamometry, UEB uterine electrogram burst, AR activation ratio, EMMI electromyometrial imaging, MAR maximal activation ratio, ACS activation curve slope, FAR fundal early activation ratio. Source data are provided as a Source Data file.

The detailed sequential activation process during the first contraction was demonstrated in Fig. 3c. The upper row shows the sequential uterine isochrone maps at different times. The warm-colored (red and yellow) regions activated early, the cool-colored regions (cyan and blue) activated late, and the gray-colored regions were not activated. The lower row shows the activation ratio (AR), defined as the percentage of uterine regions that were activated at times associated with each uterine map above. AR is calculated as dividing the area of the activated uterine region by the total uterine area as a function of time. At the end of the activation process, the complete isochrone map of uterine activation (Fig. 3d) was generated to visualize the electrical activation pattern during the entire uterine contraction. The isochrone map reveals a complete 3D activation sequence, which does not show clear long-distance propagation. First, EMMI can detect the active or inactive region during a contraction. When a large portion of the uterus remains inactive, there is insufficient myometrium to support long-distance propagation. Second, even when a large portion of the uterus is active, we did not find cardiac-like long-distance propagation within the activated region.

Based on the rich spatial and temporal information in the isochrone map, an EMMI activation curve can be generated to reflect the temporal change of the AR over time during the entire uterine contraction period (Fig. 3e). The morphology of the EMMI activation curve reflects multiple key features of uterine contraction. Maximal activation ratio (MAR) can be quantified as the total activated myometrium by the end of the contraction. The activation curve slope (ACS) is defined to reflect the slope of the activation process (black dashed line in Fig. 3e), defined as MAR divided by the time taken to reach MAR during a contraction. Based on the activation curve, the initial 33% of the active myometrium regions can be detected and defined as early activation regions and mapped back onto the 3D uterine surface to form the early activation map (Fig. 3f). In the early activation map, early active regions were shown in red, late active regions were shown in blue, and inactive regions were shown in gray (Fig. 3f). The fundus area (the 25% of the uterine surface area in the anatomical superior uterine segment, see Method) was labeled by the white dashed line (Fig. 3f), and the fundal early activation ratio (FAR) is defined as the percentage of early activation region located within the fundus area. FAR measures the extent of fundal myometrium involved in the early activation during a contraction.

Imaging uterine contractions during labor in nulliparous patients

EMMI was employed to study five nulliparous subjects (Subjects #1–5) in the active phase of term labor (Fig. 4). Subject #1 was imaged by EMMI when her cervical dilation was 3.5~4 cm (Fig. 4a). The prominent activation feature of the isochrone maps is that the activated myometrial regions were small (the gray indicates the inactive myometrium. MAR: 6.25%, 8.13%, and 19.38%) and primarily distributed at the middle and lower segments of the uterus. Based on the isochrone maps, the uterine activation curves were derived (blue curves in Fig. 4a; see details in Fig. 2e). For Subject #1, the uterine contraction activation curves were flat. The subject’s ACS values were low (0.25%/s, 0.22%/s, and 0.34%/s), and FAR values were zeros. The EMMI isochrone maps and indices suggested that the subject’s uterus was not yet electrically mature nor strongly engaged in generating forceful, synchronized contractions during the period of the electrical recording. It took 7.01 h for the subject to reach full cervix dilation after the electrical recording, and the average cervical dilation rates were 0.86 cm per h. Combined with the subject’s clinical data, EMMI findings suggest that uterine contractions involve a small amount of myometrium, as indicated by the low MAR, at the early stage of active labor.

a In Subject #1, the cervical dilation changed from 3.5 cm to 4 cm during the electrical recording and the cervix fully dilated to 10 cm 7.01 h after the recording was completed. The cervical dilation rate was calculated at 0.86 cm per h. 3D EMMI isochrone maps for three representative contractions were shown in four views. The uterine regions in warm colors were activated earlier, cool colors regions activated later, and gray regions were inactivated. The color bar on the left denotes the activation time. The activation curve and associated EMMI indices were derived from each contraction. The early activation map highlighted the early activations (earliest 33% percent of activation, red), and the fundal area was labeled by a white dashed line. b–e Results for Subjects #2, #3, #4, and #5. Same format with Subject #1. 3D three-dimensional, EMMI electromyometrial imaging, CD cervical dilation, MAR maximal activation ratio, ACS activation curve slope, FAR fundal early activation ratio. Source data are provided as a Source Data file.

For Subject #2 (Fig. 4b), three representative contractions were imaged at 4.07 h before full cervix dilation. The subject’s cervical dilation remained at 4 cm unchanged throughout the electrical recording. Although the entire uterus was not activated during the contractions, as shown in the isochrone maps, the activated uterine regions are much greater than in Subject #1. The MAR has a higher value than those in the first subject, suggesting a higher value of MAR (38.75%, 48.44%, 50.31%). Interestingly, the MAR is increasing across different contractions during the electrical recording period, suggesting that the uterus is actively recruiting more myometrium (e.g., 12.56% more myometrium was recruited into the third contraction than in the first). Within the activated uterine regions, different activation sequences were imaged for the three contractions, and no fixed early activation regions were observed. However, all three contractions were activated from the anterior-inferior fundus and lateral areas of the uterus (red-yellow regions). Activation curve slope (ACS) also increased from 1.19%/s to 1.37%/s, representing a 0.18%/s increase, suggesting the active uterine regions contracted in a more synchronized fashion in the later contractions. Fundal early activation ratios (FAR) were 7.5%, 1.25%, and 36.25%, respectively, suggesting a more fundus-initiated uterine contraction in the third contraction.

The cervical dilation of Subject #3 (Fig. 4c) was maintained at 5 cm during the electrical recording, which is 1 cm larger than that in Subject #2 (Fig. 4b). Three representative contractions were imaged at 14.41 h before full cervix dilation. However, the time to full dilation was 14.41 h, and the cervical dilation rate was 0.17 cm per h, indicating that this subject was in the latent phase of labor and had slow clinical labor progress. Similarly, we did not observe identical activation patterns and the existence of consistent initiation sites in Subject #2. Specifically, Subject #3 has high and increasing MAR for the three contractions (51.86%, 58.39%, and 65.22%). Similarly, the ACS also increased dramatically in the last two contractions (1.19%/s, 1.74%/s, and 1.75%/s). It was also noted that the early activation region (red-yellow) was dominantly located in the fundus region for the three contractions (FAR: 33.75%, 46.25%, and 31.25%), suggesting fundus-initiated contractions in this subject during the electrical recording. Despite the strong uterine contractions, the subject’s cervix dilated at a very slow rate of 0.17 cm per h to the full cervical dilation after the electrical recording.

EMMI was used to examine two subjects with cervical dilation greater than 5 cm (Subjects #4 and #5). Subject #4 (Fig. 4d) was mapped at 6.5~9.5 cm cervical dilation and 0.23 h before full dilation, dilated at the rate of 2.17 cm per h, which was much faster than that observed in Patient #1–3. The uterus was highly active (MAR: 96.88%, 71.56%, and 83.13%), fairly synchronized (ACS: 2.15%/s, 2.00%/s, and 1.99%/s). and had high FAR (40%, 67.50%, and 56.52%). The MAR values of those contractions are much higher. Similar observations were made for Subject #5 (Fig. 4e). Considering that both Subjects #4 and #5 have strong uterine contractions suggested by high MAR, ACS, and FAR, the significant difference in cervical dilation rates between the two subjects suggested an inter-subjects difference in cervix properties as we observed for subjects earlier in active labor (Fig. 4b, c).

Imaging uterine contractions during labor in multiparous patients

Five multiparous subjects were imaged by EMMI (Fig. 5). Similar to the findings in the nulliparous subjects, no fixed initiation sites or consistent activation patterns during uterine contractions were observed. In contrast to the uterine contractions imaged at the early stage of active labor in nulliparous subjects (Fig. 4a, b), EMMI found larger MAR, ACS, and FAR values in the uterine contractions from multiparous subjects (Fig. 5a, b). Subject #6 was mapped at 4 cm cervical dilation, 6.95 h before full dilation, and progressed at 0.86 cm per h after mapping. The MAR (38.75%, 52.50%, and 73.44%) were high and increasing. The ACS (0.91%/s, 1.20%/s, and 2.11%/s) and FAR (6.25%, 12.50%, and 58.75%) followed the same trend as the MAR, indicating the subject was experiencing uterine contractions with quickly increasing strength during the period of the electrical recording. In Subject #7, the cervical dilation range was 4~4.5 cm, and the time to full dilation was 1.62 h. Her labor progressed at dilation rates of 3.39 cm per h. The MAR (50.93, 40.68%, and 42.86%), ACS (2.37%/s, 1.41%/s, and 1.02%/s), and FAR (36.25%, 3.75%, and 15.00%) were high in this subject. In the early stage of active labor, EMMI found that the uterine contractions in the multiparous subjects seem stronger than those in the nulliparous subjects, which may suggest earlier electrical maturation.

a In Subject #6, the cervical dilation remained at 4 cm and reached 10 cm in 6.95 h after the EMMI recording was completed. b–e Results for Subjects #7, #8, #9, and #10. Same format with Subject #6. 3D three-dimensional, EMMI electromyometrial imaging, CD cervical dilation, MAR maximal activation ratio, ACS activation curve slope, FAR fundal early activation ratio. Source data are provided as a Source Data file.

At the later stage of active labor in the multiparous subjects (Fig. 5c–e), the MAR, ACS, and FAR values were not significantly increased compared to the nulliparous subjects (Fig. 4c–e). Subject #8 was mapped at 5 cm cervical dilation, 3.47 h before full dilation, and progressed at 1.44 cm per h after mapping. The MAR (20.94%, 25.63%, and 25.31%) was <30%. The ACS (0.43%/s, 0.43%/s, and 0.49%/s) were small and FAR (16.25%, 21.25%, and 8.75%) were normal. In Subjects #9 and #10, labor progressed at dilation rates of 2.47 cm per h and 3.52 cm per h, which are ~2.5 and 3.4 times faster than the average rate of 1 cm per h for 90% of the population. The cervical dilation ranges were 5~6.5 cm and 6.5~8 cm, and the time to full dilation was 1.42 h and 0.57 h, respectively. The MAR (15.63%, 19.38%, and 35.63%; 10.31%, 26.25%, and 31.25%) was <40% in both cases. The ACS (0.40%/s, 0.66%/s, and 1.09%/s, 0.26%/s, 0.54%/s, and 0.46%/s) was small and the FAR (6.25%, 7.50%, and 35%, 10.00%, 21.25%, and 31.25%) were normal. These findings may suggest that uterine contractions with lower MAR at the later stage of active labor are sufficient to remodel the cervix effectively and rapidly in the multiparous subjects.