Detection of obstetric anal sphincter injuries using machine learning-assisted impedance spectroscopy: a prospective, comparative, multicentre clinical study
=============================================================================================================================================================

* Katarzyna Borycka
* Marcel Młyńczak
* Maciej Rosoł
* Kacper Korzeniewski
* Piotr Iwanowski
* Hynek Heřman
* Petr Janku
* Małgorzata Uchman-Musielak
* Erik Dosedla
* Enrique Gonzalez Diaz
* Iwona Sudoł-Szopińska
* Michał Mik
* Carlo Ratto
* Antonino Spinelli

## Abstract

**Objective** To evaluate the clinical performance and safety of the ONIRY system for obstetric anal sphincter injuries (OASI) detection versus three-dimensional endoanal ultrasound (EAUS).

**Design** A prospective, comparative, multicentre, international study. Setting: Poland, Czechia, Slovakia, and Spain.

**Population** 152 women between the first moments up to 8 weeks after vaginal delivery.

**Methods** Participants underwent EAUS and were allocated to groups based on OASIS classification: A (no perineal tear), B (1st or 2nd degree tear), or C (3rd or 4th degree, anal sphincters affected). Electric impedance was measured in the anal canal using the ONIRY system. The primary endpoint was the diagnostic outcome of impedance spectroscopy versus EAUS. Adverse events were collected. Part II involved in silico modelling and 10-time 10-fold cross-validation for automated analysis.

**Main Outcome Measures** Accuracy, sensitivity, and specificity.

**Results** 30 women were allocated to group A, 61 to group B, and 61 to group C. The diagnostic outcome was determined for 147 participants. The accuracy, sensitivity, and specificity of the ML-assisted impedance spectroscopy were 87.0 ± 0.5%, 90.6 ± 2.0%, and 84.6 ± 1.9%, respectively, compared with EAUS. After data cleaning, the performance metrics of the proposed final ML model for ONIRY were: 90.0 ± 0.4%, 90.0 ± 1.2%, and 90.0 ± 0.7%, respectively. No adverse device effects or deficiencies were observed.

**Conclusion** The ML-assisted impedance spectroscopy appears to be a high-performance and safe method for rapid OASI detection, complementing digital rectal examination in obstetric settings.

**Funding** The study was financed by the Polish National Centre for Research and Development (POIR.01.01.01-00-0726/18).

Keywords
*   obstetric anal sphincter injury (OASI)
*   anal sphincter
*   perineal tear
*   faecal incontinence
*   rapid diagnostics
*   clinical study
*   impedance spectroscopy
*   machine learning

## 1. Introduction

Obstetric anal sphincter injuries (OASIs) represent a frequent complication of vaginal delivery, directly implicated in various degrees of continence impairment or frank faecal incontinence (FI). It is estimated that every fourth woman having given vaginal delivery suffers from some level of OASI1–9, translating to a 50-60% risk of developing FI either immediately following delivery10 or later in her life11–12. FI has been reported to have a destructive impact on women’s quality of life, affecting their social interactions, professional activities, family dynamics, and intimate relationships13–14.

There is still not acceptably available diagnostic and screening tool worldwide for detection of OASI in the early postpartum period. So far, obstetricians must consider this challenging diagnosis by performing the digital rectal examination as the only available diagnostic option in practice as a rule. This can be though troublesome, especially for inexperienced obstetricians and midwives. This is why the rate of primarily undetected OASIs still reaches as much as 80%15–17.

The relatively high prevalence of FI observed in women with undetected OASI following vaginal delivery12,18–19 prompts clinicians to revise the algorithms of obstetric care. Still, the gold standard for advanced diagnostics of anal sphincters lesions is the endoanal ultrasound (EAUS), a well-established and well tolerated technique with almost 100% sensitivity in identifying structural abnormalities of the anal sphincters20. As found out in a Cochrane Library review, the use of EAUS prior to obstetric perineal repair is associated with a reduction in the rate of severe FI at 6 months21. However, EAUS is of limited availability at maternity care facilities, which, admittedly, may be further confined by the lack of experienced diagnosticians available there22. As the interpretation of the EAUS imaging requires expertise, its actual performance does not reach 100% so a proportion of women is misdiagnosed with OASI even with the use of this method12,23–24.

Several studies on transperineal ultrasound (TPUS) have been reported over the last decade as a variation of ultrasound imaging of the anal sphincter that demonstrate a good correlation with EAUS results for the diagnosis of OASI25–27. A study in which TPUS was performed immediately after delivery and prior to sphincter suturing (if OASI detected) showed that more OASIs were identified on TPUS compared with digital rectal examination28. TPUS is not, however, recognised as a reference method for OASI detection.

Driven by the clinical need, the challenge of inventing and developing a novel device to detect anal sphincter injuries utilising electric impedance spectroscopy technique was taken up as the latter is an established physical marker of biostructure/tissue condition29–30, already successfully used in other medical domains31–34. With this technique, a sinusoidal electrical current, below the sensation threshold, is locally applied to the human body at different frequencies and the voltage response is measured from which impedance is calculated. So far, to the authors’ knowledge, there has been no experience with utilising this technique for diagnostics of the perianal area. An investigator-initiated proof-of-concept research study with 22 patients, followed by two pilot clinical studies with prototype devices (see below) in a total of 69 post-partum women, have been completed and previously reported wherein the method validity was demonstrated35–38. However, it was found out that due to subtility of differences in raw impedance parameters between normal (non-injured) and injured tissue structures, clinically useful sensitivity and specificity could only be achieved by utilising nonlinear machine learning (ML) algorithms for ultimate test outcome interpretation. Prototypes of such device, called the ONIRY system, with the ML module gradually trained with the clinical data collected, were designed and developed35–36. The overall concept for this system was to serve as a rapid detection method for OASI that could be applied immediately after delivery.

This clinical study was aimed at evaluating clinical performance and safety of the ONIRY system for rapid detection of OASI versus EAUS as the reference diagnostic method following re-training of the ML model with data from a larger, balanced (with and without OASI) population of post-partum women.

## 2. Materials and Methods

A prospective, comparative, multicenter, international clinical study was designed, composed of two parts: the clinical conduct (Part I) and modelling and ML (Part II). Part I of the study was conducted from 2021 to 2022 at five European centers in Czech Republic, Slovakia, Poland, and Spain. The study design and conduct were in line with the Good Clinical Practice guidelines for medical device studies (ISO 14155:2020). All approvals by the national regulatory authorities as well as positive opinions by the ethics committees, per local regulations, were obtained prior to study initiation. Written informed consent was collected from each study participant before enrolment. The study was registered at [ClinicalTrials.gov](http://ClinicalTrials.gov) under [NCT04903977](http://medrxiv.org/lookup/external-ref?link_type=CLINTRIALGOV&access_num=NCT04903977&atom=%2Fmedrxiv%2Fearly%2F2024%2F06%2F04%2F2024.06.03.24308379.atom).

### Study design

The study was designed to enroll a total of approximately 150 women between 18 and 49 years old, primiparous, or multiparous, from the first moments up to 8 weeks after vaginal (spontaneous or assisted) delivery of a singleton, live fetus, in any presentation, in gestational week 34 or higher. All inclusion and exclusion criteria are listed in **Supplementary File S1**, along with the detailed study plan.

Three study groups were pre-defined, with fixed numbers of participants enrolled as to ensure generation of balanced data from women with and without OASI. Participants were initially enrolled in these groups based on the evaluation made immediately after delivery according to the 4-degree perineal tears scale39: approximately 30 women were planned for group A (no visible perineal tear), approximately 60 for group B (clinically detectable first- or second-degree perineal tear, including episiotomy), and approximately 60 for group C (clinically detectable third- or fourth-degree perineal tear, affecting anal sphincters, regardless of primary repair).

The study duration for each participant was from 2 days up to 5 weeks and included 3 study visits. At the first visit, performed from the first moments up to 8 weeks after delivery, collection of medical history, including pregnancies and births, a physical examination, including proctological and gynaecological assessment, a 12-lead electrocardiogram (ECG) recording, and clinical signs of FI using the Wexner score were carried out. At the same visit, three-dimensional (3-D) EAUS, used as the reference diagnostic method and a tool for the ultimate study group allocation, was performed. Following such allocation (per EAUS-based OASIS classification40–41), the impedance spectroscopy by ONIRY system was performed at the second study visit (on the same day as the first study visit of up to 7 days after it). A web-based application was utilised for providing the preliminary test interpretation (only to give the operator an experience of getting the ONIRY examination result of OASI detected or not detected); since this interpretation was generated based on the ML model trained on data from the previous, small pilot clinical studies, it could differ from the ultimate interpretation obtained using the ultimate ML model trained within Part II of this study. For the exploratory analysis of reproducibility of impedance measurements, two consecutive measurement runs were performed per patient and an arbitrary criterion for discrepancy was applied, i.e., a difference between the first and second measurement runs greater than 1 kΩ for the frequency equal to 1 kHz. A 12-lead ECG was repeated immediately after impedance measurement with ONIRY system.

Faecal calprotectin levels were also measured as to assess any correlation with the results of tissue electrical impedance. At the third study visit, from 0 up to 28 days after the second visit, anal sphincter function was evaluated using a high-resolution anorectal manometry (an optional procedure, subject to actual availability at the given study site). Vital signs were assessed at each study visit.

No blinding of the study was deemed feasible. However, it was not considered as required for securing objectivity of the study outcomes as the preliminary interpretation of the impedance measurement displayed by the ONIRY system immediately after the impedance measurement (generated based on the ML model established from the previous, small pilot clinical studies) was independent from the test operator and thus not impacted by the EAUS result knowledge at study entry. Furthermore, as to minimise any bias related to the EAUS evaluation as the primary reference method, a specific technical control measure was in place that allowed for performance of the ONIRY examination only after the EAUS result and interpretation were fixed (entered in the electronic Case Report Form).

#### Study endpoints

The diagnostic outcome of ONIRY examination versus 3-D EAUS assessed per the OASIS classification was set as primary endpoint and used for the conclusion on the diagnostic performance of the ONIRY system (following the application of the ML algorithms re-trained in Part II of the study).

For the EAUS-based Diagnostic Outcome, OASI was considered detected as long as any depth, length, or circumference range of either anal sphincter (external or internal) was captured (score >2 by OASIS classification).

For the exploratory analysis of reproducibility of impedance measurements, the discrepancies were considered significant if the impedance difference between the first and second ONIRY measurement runs were greater than 1 kΩ for the frequency equal to 1 kHz.

Results obtained for secondary endpoints related to diagnostic performance assessed in this study, i.e., the Diagnostic Outcomes with other diagnostic measures utilised as reference methods for construction of the respective ML models, i.e., the digital rectal examination and high-resolution anorectal manometry, are not included in this report.

Adverse Events (AE) were recorded in each participant from enrolment to the last study visit.

### ONIRY system

The impedance spectroscopy examination was performed in each study participant using the ONIRY system composed of 3 elements: the impedance spectrometer, the anal probe (ONIRY Probe), and the ML module. The spectrometer generates the sinusoidal application current in the frequency range of 1-100 kHz with an amplitude below the threshold of sensation and pain.

The ONIRY Probe, made of a biocompatible, rigid, plastic material, has a diameter of 12 mm (at the electrodes site; the head has a diameter of 19 mm at the highest), and is composed of 8 stainless steel electrodes allowing for the measurement of the impedance module, phase shift, resistance, and reactance in perianal area. The ONIRY Probe has a handle with a marker indicating the position to be placed in the anal canal (see Figure 1).

![Figure 1.](http://medrxiv.org/https://www.medrxiv.org/content/medrxiv/early/2024/06/04/2024.06.03.24308379/F1.medium.gif)

[Figure 1.](http://medrxiv.org/content/early/2024/06/04/2024.06.03.24308379/F1)

Figure 1. ONIRY Probe with the marker location.

The ONIRY Probe is inserted into the anal canal for approximately 1 minute. It should not be rotated or moved during the measurement. The examination is carried out in supine position (or in the patient lying on her left side with knees flexed, as per operator preference) as presented in Figure 2.

![Figure 2.](http://medrxiv.org/https://www.medrxiv.org/content/medrxiv/early/2024/06/04/2024.06.03.24308379/F2.medium.gif)

[Figure 2.](http://medrxiv.org/content/early/2024/06/04/2024.06.03.24308379/F2)

Figure 2. The participant’s position during the ONIRY examination (a) along with position of the probe marker (b).

Based on raw electric impedance data collected by the anal probe and read by the ONIRY Spectrometer, the statistical parameters for the frequency sub-compartments are determined and used as the input vector for the ML model. Following the impedance measurement, with such ML support, the ONIRY system determines whether the patient has OASI and returns “PASS” (no OASI detected) or “REFER” (OASI detected). The algorithm also prevents from conducting the measurements in case abnormal use of the ONIRY Probe is detected (impedance data compatible with an improper insertion in the anal canal or no insertion at all).

### Endoanal ultrasound

The imaging of the external and internal anal sphincters, as well as the puborectalis muscle, was performed. The 3-D EAUS examination conditions were standardized (a study-specific EAUS manual was implemented at all study centres). The 3-D EAUS findings were interpreted according to the OASIS classification41–42 for the primary endpoint. In order to evaluate if evaluation according to other semiquantitative scores established for EAUS, the Starck scale42 or the Norderval scale43, could positively impact the diagnostic performance of ONIRY system once such scores are used instead of OASIS classification, each EAUS examination was additionally rated by its operator using these two scores (and respective ML models were constructed for further evaluation within exploratory analyses).

### Statistical analysis

For calculation of diagnostic performance metrics for ONIRY, the Diagnostic Outcome was determined separately for each performance endpoint, as: 

1.  Diagnostic Success: presence (True Positive) or absence (True Negative) of OASI consistently detected by the ONIRY examination and the reference diagnostic method, or

2.  Diagnostic Failure: mismatch (False Positive or False Negative) of the OASI detection by the ONIRY examination and the reference diagnostic method, or

3.  Diagnostic Indeterminate: no ONIRY or reference diagnostic method result available or interpretable.

For the primary endpoint, 3-D EAUS result (by OASIS classification) served as the reference method. Evaluation of the exploratory endpoints (diagnostic outcome with 3-D EAUS evaluated with Norderval44 or Starck43 scales) was performed accordingly. Accuracies were defined as Diagnostic Successes/Total, sensitivities as True Positives/(True positives + False Negatives), and specificities as True Negatives/(True Negatives + False Positives). Also, the F1 score, and Matthew’s Correlation Coefficient (MCC)44 were calculated.

The safety profile of ONIRY system was evaluated using descriptive statistics.

### Data analysis (Part I) **and** Machine Learning modelling (Part II)

Details for the algorithm utilised for interpretation of the ONIRY examination results presented by ONIRY in Part I of the study, with the preliminary ML model based on data from the two previously completed pilot clinical studies, are included in **Supplementary File S2.**

Details of the *in silico* analyses conducted within Part II of the study, with exploratory data analysis dimensionality reduction, ML modelling, and ultimate performance analysis with 10-fold cross-validations, are included in **Supplementary File S3**.

Within the per-patient reproducibility analysis, in case of a significant discrepancy between the two ONIRY measurement runs recorded from the same participant, the second measurement run was removed from the analysis dataset for the final algorithm for the ONIRY system (of note, in the real-world use, just one ONIRY measurement run will be conducted).

## 3. Results

### Study population

One hundred fifty-three participants were screened and 152 enrolled (1 screening failure). Following performance of the 3-D EAUS with OASIS classification, the ultimate allocation was: 30 in group A (no visible perineal tear and no OASI), 61 in group B (clinically detectable first- or second-degree perineal tear but no OASI), and 61 in group C (clinically detectable third- or fourth-degree perineal tear with OASI).

The key characteristics of the study population, per study group and in total, are presented in Table 1.

View this table:
[Table 1.](http://medrxiv.org/content/early/2024/06/04/2024.06.03.24308379/T1)

Table 1. Characteristics of the study population.

Fifteen participants prematurely discontinued the study: 3 (10.0% (3/10)) in group A, 10 (16.4% (10/61)) in group B, and 2 (3.3% (2/61)) in group C; all discontinuations occurred between the second and last visits.

### Part I (Clinical conduct)

All participants enrolled were evaluable for the primary endpoint as well as safety. Diagnostic Success or Failure could be determined for 147 participants for the primary endpoint. For the remaining 5 participants (enrolled at 3 clinical sites), the Diagnostic Outcome was Indeterminate due to electric impedance measurements highly outlying from the measurement range of the ONIRY Spectrometer, not corresponding with the impedance of the tissues measured in the anal canal. Consequently, the Diagnostic Success or Failure were assessable for 60 participants with OASI (group C) and 87 participants with no OASI (groups A and B) for the primary endpoint.

Using the limited ML model trained based on data from previous small (pilot) clinical studies with ONIRY, the diagnostic performance metrics were of 66.7% for sensitivity, 57.5% for specificity, 61.2% for accuracy, 0.58 for F1 score, and 0.24 for MCC (see **Supplementary File S4** for details).

No significant correlations between faecal calprotectin levels and results of EAUS or ONIRY examinations were found.

### Part II (*In silico* Machine Learning modelling)

The ultimate performance analysis was performed using the ML model trained with data generated in Part I of this study (raw electric impedance data by ONIRY and data on OASI detection by EAUS as the reference). A total of 298 impedance measurements (meeting quality requirements, see **Supplementary File S5**) were included, i.e., 122 corresponding with OASI (group C participants) and 176 with no OASI (group A and B participants). Following the exploratory data analysis, dimensionality reduction, and modelling with the artificial neural networks, markedly improved diagnostic metrics (with sensitivity slightly above 90% and specificity slightly below 85%) were observed, as presented in Table 2 (see **Supplementary File S6** for individual statistics for each single 10-fold cross-validation).

View this table:
[Table 2.](http://medrxiv.org/content/early/2024/06/04/2024.06.03.24308379/T2)

Table 2. Mean performance metrics of the ONIRY system using the models trained on data from Part I of the study, in the assessment relative to 3-D Endoanal Ultrasound and OASIS classification.

The exploratory analysis of per-patient reproducibility of impedance measurements recorded in Part I of the study showed discrepancies between the two impedance measurement runs in 19 participants (12.8%).

The exploratory analysis of performance assessment in reference to 3-D EAUS with the use of the classifications of anal sphincter injuries according to Starck43 and Norderval44(also utilised to train alternative artificial network ML models, respectively) resulted in 83.8 % for sensitivity, 88.1 % for specificity, 86.4 % for accuracy, 0.83 for F1 score, and 0.73 for MCC for Starck scoring and 84.3 % for sensitivity, 89.3 % for specificity, 87.4 % for accuracy, 0.84 for F1 score, and 0.74 for MCC for Norderval scoring.

### Safety results

No deaths, serious AEs, or AEs leading to premature withdrawal from the study were reported during the study conduct. A total of 22 AEs were observed in 21 participants (4 AEs in group A, 5 in group B, and 13 in group C); all these AEs were reported posteriorly to the ONIRY examinations. Four AEs occurred more than in one participant, of which the most frequent were nasopharyngitis (2.6%, 4/152) and COVID-19 (2.6%, 4/152).

No AE was considered related to the use of the ONIRY system (i.e., no adverse device effect was reported).

No trend was observed in ECG for any parameter when comparing the baseline (recordings prior to the ONIRY examination) with post-ONIRY application recordings. No post-ONIRY application clinically significant abnormality recorded as an AE within the cardiovascular system, either.

### Post-hoc analysis

Following the exploratory analyses of per-patient reproducibility of impedance measurements 19 measurement files (those generated from the second measurement in each case) were eliminated for the production version of the ultimate ML model for the ONIRY system, leaving 93.6% (279/298) of the dataset for the analysis (“limited dataset”). The limited dataset included 279 impedance measurements, i.e., 117 corresponding with OASI and 162 with no OASI. A post-hoc performance analysis with cross-validations performed for this ML model showed the diagnostic metrics with either sensitivity and specificity of ONIRY system at 90.0%, as presented in Table 3 (see **Supplementary File S6** for individual statistics for each single 10-fold cross-validation).

View this table:
[Table 3.](http://medrxiv.org/content/early/2024/06/04/2024.06.03.24308379/T3)

Table 3. Mean performance metrics of the ONIRY system using the Machine Learning model trained based on the limited clinical study dataset in the assessment relative to 3-D Endoanal Ultrasound and OASIS classification.

## 4. Discussion

### Key Findings

The ONIRY system showed high diagnostic performance in detection of OASI versus EAUS results as the reference method (per OASIS classification) in the study population of women after vaginal delivery, enriched for the cases with OASI (40.1%, 61/152 of all women enrolled). As a result of Part II of the study, following the *in silico* ML model re-training with the data generated in Part I of the study, an accuracy of 87.0 ± 0.5%, sensitivity of 90.6 ± 2.0%, and specificity of 84.6% ± 1.9% were observed for ONIRY. After limiting the dataset (by 6.4%, 19/298) based on the discrepancies in the consecutive impedance measurements, the next ML model, to be utilised for the final ONIRY system, reached an accuracy of 90.0 ± 0.4%, with a sensitivity of 90.0 ± 1.2% and a specificity of 90.0 ± 0.7%. This demonstrates the diagnostic benefit of the ML support for analysis of electric impedance data for OASI detection achieved thanks to re-training the ML model (models) with a relatively large and well-balanced (OASI vs. no OASI) dataset. This is in contrast with the intermediary moderate performance (accuracy of 61.2%) of the ML-assisted impedance spectroscopy method observed in Part I of the study for the same setting (same diagnostic reference method and study participants) but utilising the ML model constructed solely with the data from the smaller (pilot), poorly balanced (OASI cases underrepresented) previous clinical studies. The performance improvement of the ultimate ML model constructed for the ONIRY system appears to be due to a higher variability of raw impedance data collected in this clinical study compared with the data from the previous (pilot) clinical studies with ONIRY as well as the dataset balance (OASI cases highly represented).

The alternative performance metrics obtained when utilising the exploratory reference methods (OASI evaluated by EAUS using Starck and Norderval scores, including alternative ML models constructed) did not demonstrate apparent benefits over the primary performance measure as presented above.

No safety concern has been identified with the use of the ONIRY system.

### Clinical Implications

Currently, no rapid, easy-to-use, reliable diagnostic tool is available in maternity care settings that could be used immediately or shortly (within hours) after a vaginal delivery and operated by all personnel within the obstetric team. As demonstrated in this study, the ONIRY system, providing a straightforward interpretation of the electric impedance results in the anal canal, can detect OASI with high accuracy, at a level of approximately 90% (compared with 3-D EAUS, per OASIS classification, utilised as the reference method). With such metrics, this method has the potential to be adopted as a method for detection of OASI effectively complementing the digital rectal examination, especially where the 3-D EAUS is not or scarcely available for such primary detection. Since a single measurement run of the ONIRY system (including the automated analysis and presentation of the results) takes less than a minute, it may have an acceptance potential for medical (obstetric) practice.

As found out by a Cochrane Library systematic review, EAUS performed immediately after delivery, before any perineal repair, allows for an increase in OASI detection and thus the primary sphincter repair rate. Also, EAUS completed prior to the perineal repair may be associated with a reduced risk of developing severe FI21. This observation indicates that obstetricians, frequently having no efficient tool for OASI detection in their routine practice, may not be able to timely initiate the optimal treatment. As evidenced in literature, training in perineal anatomy and detection of pelvic floor injuries is still insufficient among obstetricians and midwives45–48.

What favours EAUS as the diagnostic method is the possibility to separately visualise both external and internal anal sphincters. This allows for a targeted repair of the external or both internal and external sphincter. Anyway, EAUS examination is not sufficiently available, as a rule, in the obstetric settings. It is one of the main factors contributing to the underdiagnosis of OASI (missing OASI). Even if the EAUS advantages are undeniable, this procedure remains challenging from both operational and economical standpoints, depending primarily on the availability of experienced operators, as well as equipment. Therefore, there is an unmet need for establishing an effective, non- or lowly invasive method for OASI detection for maternity care facilities and midwife practices that have no reliable diagnostic tool in place. Optimally, such a method should allow for its application as a screening test with clearly defined criteria for extended diagnostics as required and available. In case the impedance spectroscopy proves to perform in clinical practice, it might also help guide the overall postpartum management of women in the maternity care facilities and streamline early referrals to specialised surgery units.

With the ML implemented in the ONIRY system, the raw impedance data is automatically interpreted, and the ultimate test result presented in a simple binary (injured/non-injured or pass/refer) form. Such a feature of the ONIRY examination would presumably enable its routine performance even by an operator with limited experience.

No risks or safety concerns have been identified for the electric impedance-based method for OASI detection, in particular with the use of the ONIRY system. From the performance standpoint, the following contraindications for ONIRY examination could be considered: any implants in the pelvic area, major malformations of the perianal area, or coincident connection with any electronic medical or surgical equipment or device generating alternating current above 1 kHz.

## Limitations and research implications

There are several limitations of this study.

The study population was specifically selected, i.e., enriched with OASI cases (group C, n=61). Such imbalance was deemed necessary to improve the internal balance of the dataset as to increase the robustness of the ML model (models) constructed due to expectedly higher internal data coherence. In fact, it is a common approach in the development and training of artificial intelligence/ML tools, applicable not only in the medical field but also beyond49.

Moreover, the primary endpoint set was surrogate (OASI detection) and not ultimately clinically meaningful (which could be the FI or a quality of life-based endpoint). This surrogate endpoint was set to facilitate a comparison of the ONIRY system with the 3-D EAUS as the gold diagnostic standard, feasible within a relatively short time span. Establishing a clinically meaningful benefit for the ONIRY system would require further long-term follow-up studies spanning rather years than months or weeks considering the dynamics of FI development over time in women with OASI. Such putative benefit of rapid detections of OASI would be, however, highly dependent on the actual therapeutic measures that are feasible and actually performed in the given obstetric settings. Of paramount importance will be the ability to quickly perform the primary repair of the anal sphincter in women with OASI detected, considering the narrow time window for such procedure, up to just 8 to 12 hours following delivery50–51.

Furthermore, for the majority (52 out of 61) of participants in group C with the sphincter primary repair made, such repair had been performed prior to enrolment in the study. This may have altered the impedance spectroscopy readouts towards non-injured (no OASI) cases. However, since the ML models, particularly the ultimate ML model, were trained on such data, it can be speculated that the real-world performance of the ONIRY system in the target population (i.e., prior to performing the primary repair) would be even greater. This is because the biostructure and geometry of the anal sphincter muscle injured and newly repaired (primary repair) is basically closer to normal than for the muscle newly injured and unrepaired (not yet repaired).

This high proportion of OASI cases with primary repair already made prior to ONIRY (and 3-D EAUS) evaluations corresponds with the difference between the study group characteristics for the time elapsed between the index delivery and ONIRY examination (median time of 28 days in group C versus 2 and 11 days in groups A and B, respectively). This imbalance reflects the enrolment difficulty the investigators and study sites had with enriching the study population with participants with 3rd and 4th degree perineal tears (i.e., the study group C), and the long diagnostic and therapeutic pathway currently being experienced by patients with OASI.

On the other hand, the inclusion of a considerable proportion of women with OASI only after the primary repair in group C allows to address, as described in the literature, clinical situations where under-classified OASIs remain unattended or ineffectively attended, despite primary repair being performed (incomplete repair)52,53. Literature data shows that the primary repair itself is performed ineffectively in over 30% of cases, mainly due to limited experience of the operator but also to the emergency nature of the repair procedure itself, usually performed during on-call hours54. It is therefore postulated that the classification of cases with the repair made shortly prior to study enrolment to the ’injured’ group is methodologically acceptable. Finally, for multiparous women, no detailed data was captured (in particular, no active diagnostics was performed) on potential undetected OASIs that would have possibly occurred at previous delivery/deliveries. In such a case, theoretically, the impedance parameters might appear normal (as the tissue had likely healed) whereas EAUS might still show some abnormality. However, an alternative diagnostic scenario might also be speculated for multiparas: an abnormal impedance spectroscopy with normal EAUS, compatible with fibrosis following an old injury.

## 5. Conclusions

This study demonstrated that the ONIRY system utilising ML-assisted impedance spectroscopy with a dedicated anal probe is a safe and effective method for rapid detection of OASI in women having given vaginal delivery.

## Supporting information

Supplementary File S1 [[supplements/308379_file03.docx]](pending:yes)

Supplementary File S2 [[supplements/308379_file04.docx]](pending:yes)

Supplementary File S3 [[supplements/308379_file05.docx]](pending:yes)

Supplementary File S4 [[supplements/308379_file06.docx]](pending:yes)

Supplementary File S5 [[supplements/308379_file07.docx]](pending:yes)

Supplementary File S6 [[supplements/308379_file08.docx]](pending:yes)

Supplementary File S7 [[supplements/308379_file09.docx]](pending:yes)

Supplementary File S8 [[supplements/308379_file10.docx]](pending:yes)

## Data Availability

Data sharing is not applicable to this article due to legal and privacy issues.

## Disclosure of interests

K.B. is a founder and management board member at OASIS Diagnostics, an author of the related patent and R&D strategy, independent consultant, and trainer of Takeda.

M.M., M.R., K.K., and P.I. are staff of OASIS Diagnostics.

A.S. is an independent consultant of Ethicon, Takeda, Pfizer, Sofar. M.UM. is an independent consultant of Regen Lab.

H.H., P.J., M.UM., E.D., and E.GD. received remuneration as a study investigator.

H.H., C.R., and A.S. are independent consultants and members of the OASIS Diagnostics’ Scientific Advisory Board.

Others declare no conflicts of interest.

## Author contributions

Conceptualization: KB, MM, PI

Data curation: KB, MM, MR, KK, PI

Formal analysis: MM, MR, KK

Funding acquisition: KB, MM

Investigation: HH, PJ, MUM, ED, EGD

Methodology: KB, MM, MR, KK, PI

Project administration: KB, MM

Resources: HH, PJ, MUM, ED, EGD

Software: MM, MR, KK

Supervision: KB, MM, PI, CR, AS

Validation: MM, PI

Visualization: MR, KK

Writing - original draft: KB, MM, MR, KK, PI

Writing - review & editing: All co-authors

## Details of ethics approval

The study was approved by the ethics committees respective for each study site: on 19 March 2021 by Ethics Committee of the Institute for Maternal and Child Care (no. 1/19.03.2021), on 27 April 2021 by Ethics Committee for Research with Medicines of the Health Areas of León and Bierzo (no. 2186), on 9 June 2021 by Ethics Committee of the University Hospital of Brno (no. 47/21Zdrav.), on 14 October 2021 by Ethics Committee at the Regional Medical Chamber in Warsaw (no. KB/1362/21) and on 25 July 2022 by Ethics Committee at AGEL Hospital Košice-Šaca (no. ONIRY 3/2/2020).

## Funding

The study was funded by the European Union as part of the Fast Track program, conducted in Poland by the National Centre for Research and Development (POIR.01.00.01-00-0726/18).

*   Received June 3, 2024.
*   Revision received June 3, 2024.
*   Accepted June 4, 2024.


*   © 2024, Posted by Cold Spring Harbor Laboratory

This pre-print is available under a Creative Commons License (Attribution 4.0 International), CC BY 4.0, as described at [http://creativecommons.org/licenses/by/4.0/](http://creativecommons.org/licenses/by/4.0/)

## References

1.  [1].Sideris M, McCaughey T, Hanrahan JG, Arroyo-Manzano D, Zamora J, Jha S et al. Risk of obstetric anal sphincter injuries (OASIS) and anal incontinence: a meta-analysis. European Journal of Obstetrics & Gynecology and Reproductive Biology 2020;252,303–312. Doi: 10.1016/j.ejogrb.2020.06.048
    
    [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1016/j.ejogrb.2020.06.048&link_type=DOI) 

2.  [2].Aigmueller T, Bader W, Beilecke K, Elenskaia K, Frudinger A, Hanzal E et al. Management of 3rd and 4th degree perineal tears after vaginal birth. German Guideline of the German Society of Gynecology and Obstetrics (AWMF Registry No. 015/079, October 2014). Geburtshilfe Frauenheilkd 2015;75(2):137–44. doi:10.1055/s-0034-1396323
    
    [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1055/s-0034-1396323&link_type=DOI) 

3.  [3].Gurol-Urganci I, Cromwell DA, Edozien LC, Mahmood TA, Adams EJ, Richmond DH et al. Third- and fourth-degree perineal tears among primiparous women in England between 2000 and 2012: time trends and risk factors. BJOG 2013;120(12):1516–25. doi: 10.1111/1471-0528.12363
    
    [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1111/1471-0528.12363&link_type=DOI) 
    
    [PubMed](http://medrxiv.org/lookup/external-ref?access_num=23834484&link_type=MED&atom=%2Fmedrxiv%2Fearly%2F2024%2F06%2F04%2F2024.06.03.24308379.atom) 

4.  [4].Dudding TC, Vaizey CJ, Kamm MA. Obstetric anal sphincter injury: incidence, risk factors, and management. Ann Surg 2008;247(2):224–37. doi:10.1097/SLA.0b013e318142cdf4
    
    [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1097/SLA.0b013e318142cdf4&link_type=DOI) 
    
    [PubMed](http://medrxiv.org/lookup/external-ref?access_num=18216527&link_type=MED&atom=%2Fmedrxiv%2Fearly%2F2024%2F06%2F04%2F2024.06.03.24308379.atom) 

5.  [5].Sultan AH, Kamm MA, Bartram CI, Hudson CN. Anal sphincter trauma during instrumental delivery. Int J Gynaecol Obstet 1993;43:263–70. doi:111/1471-0528.12363
    
    [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1016/0020-7292(93)90514-W&link_type=DOI) 
    
    [PubMed](http://medrxiv.org/lookup/external-ref?access_num=7907036&link_type=MED&atom=%2Fmedrxiv%2Fearly%2F2024%2F06%2F04%2F2024.06.03.24308379.atom) 

6.  [6].Tejedor P, Plaza J, Bodega-Quiroga I, Ortega-López M, García-Olmo D, Pastor C. The Role of Three-Dimensional Endoanal Ultrasound on Diagnosis and Classification of Sphincter Defects After Childbirth. J Surg Res. 2019;244:382–388. doi: 10.1016/j.jss.2019.06.080
    
    [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1016/j.jss.2019.06.080&link_type=DOI) 

7.  [7].Tejedor P, Bodega-Quiroga I, Plaza J, Ortega López M, Gutierrez C, García Olmo D et al. Quality of life and 3D-EUS assessment for anal incontinence after childbirth. Rev Esp Enferm Dig. 2019;111(6):453–459. doi: 10.17235/reed.2019.6040/2018
    
    [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.17235/reed.2019.6040/2018&link_type=DOI) 

8.  [8].Samarasekera DN, Bekhit MT, Wright Y, Lowndes RH, Stanley KP, Preston JP et al. Long-term anal continence and quality of life following postpartum anal sphincter injury. Colorectal Dis 2008;10(8):793–9. doi: 10.1111/j.1463-1318.2007.01445.x.
    
    [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1111/j.1463-1318.2007.01445.x&link_type=DOI) 
    
    [PubMed](http://medrxiv.org/lookup/external-ref?access_num=18266886&link_type=MED&atom=%2Fmedrxiv%2Fearly%2F2024%2F06%2F04%2F2024.06.03.24308379.atom) 
    
    [Web of Science](http://medrxiv.org/lookup/external-ref?access_num=000259307900011&link_type=ISI) 

9.  [9].Mous M, Muller SA, de Leeuw JW. Long-term effects of anal sphincter rupture during vaginal delivery: faecal incontinence and sexual complaints. BJOG. 2008;115(2):234–8. doi: 10.1111/j.1471-0528.2007.01502.
    
    [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1111/j.1471-0528.2007.01502&link_type=DOI) 
    
    [PubMed](http://medrxiv.org/lookup/external-ref?access_num=17999696&link_type=MED&atom=%2Fmedrxiv%2Fearly%2F2024%2F06%2F04%2F2024.06.03.24308379.atom) 

10. [10].Everist R, Burrell M, Mallitt KA, Parkin K, Patton V, Karantanis E Postpartum anal incontinence in women with and without obstetric anal sphincter injuries. Int Urogynecol J. 2020 Nov;31(11):2269–2275. doi: 10.1007/s00192-020-04267-8. Epub 2020 Mar 10.
    
    [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1007/s00192-020-04267-8&link_type=DOI) 

11. [11].Jangö H, Langhoff-Roos J, Rosthøj S, Sakse A Recurrent obstetric anal sphincter injury and the risk of long-term anal incontinence Am J Obstet Gynecol. 2017 Jun;216(6):610.e1-610.e8. doi: 10.1016/j.ajog.2017.02.010. Epub 2017 Feb 13.
    
    [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1016/j.ajog.2017.02.010&link_type=DOI) 

12. [12].Guzmán Rojas RA, Salvesen K Å, Volløyhaug I. Anal sphincter defects and fecal incontinence 15–24 years after first delivery: a cross-sectional study. Ultrasound Obstet Gynecol 2018; 51(5): 677–683. doi: 10.1002/uog.18827
    
    [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1002/uog.18827&link_type=DOI) 

13. [13].Meyer I, Richter HE. Impact of fecal incontinence and its treatment on quality of life in women. Women’s Health (London), 2015; 11(2): 225–238. doi: 10.2217/WHE.14.66
    
    [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.2217/WHE.14.66&link_type=DOI) 
    
    [PubMed](http://medrxiv.org/lookup/external-ref?access_num=25776296&link_type=MED&atom=%2Fmedrxiv%2Fearly%2F2024%2F06%2F04%2F2024.06.03.24308379.atom) 

14. [14].Jangö H, Langhoff-Roos J, Rosthøj S, et al. Wexner score and quality of life in women with obstetric anal sphincter injury. International Urogynecology Journal, 2020; 31(6): 1115–1121. doi: 10.1007/s00192-019-04134-1
    
    [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1007/s00192-019-04134-1&link_type=DOI) 

15. [15].Zimmo K, Laine K, Vikanes Å, Fosse E, Zimmo M, Ali H et al. Diagnosis and repair of perineal injuries: knowledge before and after expert training-a multicentre observational study among Palestinian physicians and midwives. BMJ Open. 2017;7(4):e014183. doi: 10.1136/bmjopen-2016-014183.
    
    [Abstract/FREE Full Text](http://medrxiv.org/lookup/ijlink/YTozOntzOjQ6InBhdGgiO3M6MTQ6Ii9sb29rdXAvaWpsaW5rIjtzOjU6InF1ZXJ5IjthOjQ6e3M6ODoibGlua1R5cGUiO3M6NDoiQUJTVCI7czoxMToiam91cm5hbENvZGUiO3M6NzoiYm1qb3BlbiI7czo1OiJyZXNpZCI7czoxMToiNy80L2UwMTQxODMiO3M6NDoiYXRvbSI7czo1MDoiL21lZHJ4aXYvZWFybHkvMjAyNC8wNi8wNC8yMDI0LjA2LjAzLjI0MzA4Mzc5LmF0b20iO31zOjg6ImZyYWdtZW50IjtzOjA6IiI7fQ==) 

16. [16].Andrews V, Thakar R, Sultan AH. Outcome of obstetric anal sphincter injuries (OASIS)-role of structured management. Int Urogynecol J Pelvic Floor Dysfunct 2009;20(8):973–8. doi: 10.1007/s00192-009-0883-0.
    
    [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1007/s00192-009-0883-0&link_type=DOI) 

17. [17].Ramage L, Yen C, Qiu S, Simillis C, Kontovounisios C, Tan E, Tekkis P. Does a missed obstetric anal sphincter injury at time of delivery affect short-term functional outcome? Ann R Coll Surg Engl. 2018;100(1):26–32.
    
    

18. [18].Turel FD, Langer S, Shek KL, Dietz HP. Medium- to Long-term Follow-up of Obstetric Anal Sphincter Injury. Dis Colon Rectum 2019;62(3):348–356. doi:10.1097/DCR.0000000000001297.
    
    [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1097/DCR.0000000000001297&link_type=DOI) 

19. [19].Nilsson IE, Åkervall S, Molin M, Milsom I, Gyhagen M. Symptoms of fecal incontinence two decades after no, one, or two obstetrical anal sphincter injuries. American Journal of Obstetrics and Gynecology, 2021;224(3), 276–e1.
    
    

20. [20].Albuquerque A. Endoanal ultrasonography in fecal incontinence: Current and future perspectives. World J Gastrointest Endosc 2015;7(6):575–581. doi: 10.4253/wjge.v7.i6.575.
    
    [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.4253/wjge.v7.i6.575&link_type=DOI) 

21. [21].Walsh KA and Grivell RM. Use of endoanal ultrasound for reducing the risk of complications related to anal sphincter injury after vaginal birth. Cochrane Database of Systematic Reviews 2015;(10):CD010826. doi: 10.1002/14651858.CD010826.pub2
    
    [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1002/14651858.CD010826.pub2&link_type=DOI) 

22. [22].Dietz HP. Exoanal imaging of the anal sphincters; J Ultrasound Med; 2018;37(1):263-280. doi: 10.1002/jum.14246
    
    [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1002/jum.14246&link_type=DOI) 

23. [23].Sioutis D, Thakar R, Sultan AH. Overdiagnosis and rising rate of obstetric anal sphincter injuries (OASIS): time for reappraisal. Ultrasound Obstet Gynecol 2017;50:642–647. doi: 10.1002/uog.17306
    
    [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1002/uog.17306&link_type=DOI) 

24. [24].Thomas GP, Gould LE, Casunuran F, Kumar DA. A retrospective review of 1495 patients with obstetric anal sphincter injuries referred for assessment of function and endoanal ultrasonography. Int J Colorectal Dis. 2017 Sep;32(9):1321–1325. doi: 10.1007/s00384-017-2851-3.
    
    [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1007/s00384-017-2851-3&link_type=DOI) 

25. [25].Stuart A, Ignell C, Örnö AC. Comparison of transperineal and endoanal ultrasound in detecting residual obstetric anal sphincter injury. Acta Obstet Gynecol Scand. 2019;98(12):1624–1631. doi: 10.1111/aogs.13701.
    
    [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1111/aogs.13701&link_type=DOI) 

26. [26].Ros C, Martínez-Franco E, Wozniak MM, Cassado J, Santoro GA, Elías N, López M, Palacio M, Wieczorek AP, Espuña-Pons M. Postpartum two- and three-dimensional ultrasound evaluation of anal sphincter complex in women with obstetric anal sphincter injury. Comparative Study Ultrasound Obstet Gynecol. 2017 Apr;49(4):508–514. doi: 10.1002/uog.15924.
    
    [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1002/uog.15924&link_type=DOI) 

27. [27].Ignell C, Örnö AK, Stuart A Correlations of obstetric anal sphincter injury (OASIS) grade, specific symptoms of anal incontinence, and measurements by endoanal and transperineal ultrasound. J Ultrasound 2021;24(3):261–267. doi: 10.1007/s40477-020-00485-4.
    
    [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1007/s40477-020-00485-4&link_type=DOI) 

28. [28].Wong KW, Thakar R, Sultan AH, Andrews V. Can transperineal ultrasound improve the diagnosis of obstetric anal sphincter injuries? Int Urogynecol J. 2022 Aug 2. doi: 10.1007/s00192-022-05290-7.
    
    [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1007/s00192-022-05290-7&link_type=DOI) 

29. [29].Shiffman CA, Aaron R. Low-impedance localized measurements using standard bioelectrical impedance analysis instruments. Ann NY Acad Sci 2000;904:214–217. doi: 10.1111/j.1749-6632.2000.tb06453.x
    
    [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1111/j.1749-6632.2000.tb06453.x&link_type=DOI) 
    
    [PubMed](http://medrxiv.org/lookup/external-ref?access_num=10865742&link_type=MED&atom=%2Fmedrxiv%2Fearly%2F2024%2F06%2F04%2F2024.06.03.24308379.atom) 

30. [30].Aaron R, Shiffman CA. Using localized impedance measurements to study muscle changes in injury and disease. Ann NY Acad Sci 2000;904:171–180. doi: 10.1111/j.1749-6632.2000.tb06443.x
    
    [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1111/j.1749-6632.2000.tb06443.x&link_type=DOI) 
    
    [PubMed](http://medrxiv.org/lookup/external-ref?access_num=10865732&link_type=MED&atom=%2Fmedrxiv%2Fearly%2F2024%2F06%2F04%2F2024.06.03.24308379.atom) 
    
    [Web of Science](http://medrxiv.org/lookup/external-ref?access_num=000088775500030&link_type=ISI) 

31. [31].Seward B, Rutkove MD, Aaron R, Shiffman CA. Localized bioimpedance analysis in the evaluation of neuromuscular disease. Muscle Nerve 2002;25:390–397. doi: 10.1002/mus.10048
    
    [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1002/mus.10048&link_type=DOI) 
    
    [PubMed](http://medrxiv.org/lookup/external-ref?access_num=11870716&link_type=MED&atom=%2Fmedrxiv%2Fearly%2F2024%2F06%2F04%2F2024.06.03.24308379.atom) 

32. [32].Młyńczak M, Cybulski G. Decomposition of the cardiac and respiratory components from impedance pneumography signals. Proc. of the 10th Int. Joint Conf. on Biomedical Engineering Systems and Technologies – Biodevices 2017;4:26–33. doi: 10.5220/0006107200260033
    
    [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.5220/0006107200260033&link_type=DOI) 

33. [33].Kushner RF, Schoeller DA. Estimation of total body water by bioelectrical impedance analysis. Am J Clin Nutr 1986;44(3):417–424. doi: 10.1093/ajcn/44.3.417
    
    [Abstract/FREE Full Text](http://medrxiv.org/lookup/ijlink/YTozOntzOjQ6InBhdGgiO3M6MTQ6Ii9sb29rdXAvaWpsaW5rIjtzOjU6InF1ZXJ5IjthOjQ6e3M6ODoibGlua1R5cGUiO3M6NDoiQUJTVCI7czoxMToiam91cm5hbENvZGUiO3M6NDoiYWpjbiI7czo1OiJyZXNpZCI7czo4OiI0NC8zLzQxNyI7czo0OiJhdG9tIjtzOjUwOiIvbWVkcnhpdi9lYXJseS8yMDI0LzA2LzA0LzIwMjQuMDYuMDMuMjQzMDgzNzkuYXRvbSI7fXM6ODoiZnJhZ21lbnQiO3M6MDoiIjt9) 

34. [34].Tidy JA, Brown BH, Healey TJ, Daayana S, Martin M, Prendiville W et al. Accuracy of detection of high-grade cervical intraepithelial neoplasia using electrical impedance spectroscopy with colposcopy. BJOG 2013;120(4):400–10; discussion 410-1. doi: 10.1111/1471-0528.12096
    
    [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1111/1471-0528.12096&link_type=DOI) 
    
    [PubMed](http://medrxiv.org/lookup/external-ref?access_num=23289897&link_type=MED&atom=%2Fmedrxiv%2Fearly%2F2024%2F06%2F04%2F2024.06.03.24308379.atom) 

35. [35].Młyńczak M, Borycka-Kiciak K, Uchman-Musielak M, Dziki A. Impedance Spectroscopy Method to Detect Pelvic Floor Muscle Damage - A Feasibility Study. In: World Congress on Medical Physics and Biomedical Engineering, Springer, Singapore, 2018;875–878. doi: 10.1007/978-981-10-9038-7_161.
    
    [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1007/978-981-10-9038-7_161&link_type=DOI) 

36. [36].Borycka-Kiciak K, Młyńczak M, Kiciak A, Pietrzak P, Dziki A. Non-invasive obstetric anal sphincter injury diagnostics using impedance spectroscopy. Scientific Reports 2019;9(7097):1–9. doi: 10.1038/s41598-019-43637-1.
    
    [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1038/s41598-019-55134-6&link_type=DOI) 

37. [37].Młyńczak M, Rosoł M, Spinelli A, Dziki A, Wlazlak E, Surkont G, et al. Obstetric Anal Sphincter Injury Detection Using Impedance Spectroscopy with the ONIRY Probe; Appl Sci.; 2021; 11: 637; doi: 10.3390/app11020637.
    
    [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.3390/app11020637&link_type=DOI) 

38. [38].Borycka K, Młyńczak M, Rosoł M, Iwanowski P, Uchman-Musielak M, Mik M, Sudoł-Szopińska I, Herman H, Ratto C, Dziki A, Wlaźlak E, Surkont G, Krzycka M, Pająk P, Spinelli A. Impedance spectroscopy for the diagnosis of obstetric anal sphincter injuries: the pilot experience; In: 15 Congress of the European Society of Gynecology, 2023.
    
    

39. 39.Cunningham FG, Leveno KJ, Bloom SL, Hauth J, Rouse D, Spong C. Williams obstetrics (23rd Edition), 2011. New York: McGraw-Hill Medical.
    
    

40. 40.The Management of Third- and Fourth-Degree Perineal Tears: a Green-top Guideline No 29; Royal College of Obstetrician and Gynaecologists, London; 2015
    
    

41. [41].Prevention and Management of Obstetric Lacerations at Vaginal Delivery. ACOG Practice Bulletin No. 165: American College of Obstetricians and Gynecologists’ Committee on Practice Bulletins—Obstetrics. Obstet Gynecol 2016;128(1):e1–e15. doi: 10.1097/AOG.0000000000001523
    
    [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1097/AOG.0000000000001523&link_type=DOI) 

42. [42].Starck M, Bohe M, Valentin L. Results of endosonographic imaging of the anal sphincter 2–7 days after primary repair of third or fourth-degree obstetric sphincter tears. Ultrasound Obstet Gynecol. 2003;22:609–615. doi: 10.1002/uog.920
    
    [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1002/uog.920&link_type=DOI) 
    
    [PubMed](http://medrxiv.org/lookup/external-ref?access_num=14689534&link_type=MED&atom=%2Fmedrxiv%2Fearly%2F2024%2F06%2F04%2F2024.06.03.24308379.atom) 
    
    [Web of Science](http://medrxiv.org/lookup/external-ref?access_num=000187606400011&link_type=ISI) 

43. [43].Norderval S, Markskog A, Rossaak K, Vonen B. Correlation between anal sphincter defects and anal incontinence following obstetric sphincter tears: Assessment using scoring systems for sonographic classification of defects. Ultrasound Obstet Gynecol 2008;31: 78–84. doi: 10.1002/uog.5155
    
    [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1002/uog.5155&link_type=DOI) 
    
    [PubMed](http://medrxiv.org/lookup/external-ref?access_num=18059077&link_type=MED&atom=%2Fmedrxiv%2Fearly%2F2024%2F06%2F04%2F2024.06.03.24308379.atom) 

44. [44].Chicco D, Jurman G. The advantages of the Matthews correlation coefficient (MCC) over F1 score and accuracy in binary classification evaluation. BMC genomics, 2020:21, 1–13.
    
    [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1186/s12864-020-6490-7&link_type=DOI) 
    
    [PubMed](http://medrxiv.org/lookup/external-ref?access_num=32977771&link_type=MED&atom=%2Fmedrxiv%2Fearly%2F2024%2F06%2F04%2F2024.06.03.24308379.atom) 

45. [45].De Meutter L, van Heesewijk AD, van der Woerdt-Eltink I, de Leeuw JW. Implementation of a perineal support programme for reduction of the incidence of obstetric anal sphincter injuries and the effect of non-compliance. Eur J Obstet Gynecol Reprod Biol 2018;230:119–123. doi: 10.1016/j.ejogrb.2018.09.021
    
    [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1016/j.ejogrb.2018.09.021&link_type=DOI) 
    
    [PubMed](http://medrxiv.org/lookup/external-ref?access_num=http://www.n&link_type=MED&atom=%2Fmedrxiv%2Fearly%2F2024%2F06%2F04%2F2024.06.03.24308379.atom) 

46. 46.New joint initiative to prevent severe tearing during childbirth. Royal College of Obstetricians & Gynaecologists, Royal College of Midwives: London; 2016. ([https://www.rcog.org.uk/en/news/rcog-statement-new-joint-initiative-to-prevent-severe-tearing-during-childbirth/](https://www.rcog.org.uk/en/news/rcog-statement-new-joint-initiative-to-prevent-severe-tearing-during-childbirth/)). (Accessed on 25th of January, 2022)
    
    

47. [47].NMPA Project Team: National Maternity and perinatal audit: Clinical report 2017. In. Royal College of Obstetricians & Gynaecologists. London: NHS Ligation Authority 2017.
    
    

48. [48].Bidwell P, Thakar R, Sevdalis N, Silverton L, Novis V, Hellyer A et al. A multi-centre quality improvement project to reduce the incidence of obstetric anal sphincter injury (OASI): study protocol. BMC Pregnancy Childbirth 2018 Aug 13;18(1):331. doi: 10.1186/s12884-018-1965-0
    
    [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1186/s12884-018-1965-0&link_type=DOI) 
    
    [PubMed](http://medrxiv.org/lookup/external-ref?access_num=http://www.n&link_type=MED&atom=%2Fmedrxiv%2Fearly%2F2024%2F06%2F04%2F2024.06.03.24308379.atom) 

49. [49].Batista GEAPA, Prati RC, Monard MC. A study of the behavior of several methods for balancing machine learning training data. ACM SIGKDD Explorations Newsletter. 2004;6(1):20–29. doi:10.1145/1007730.1007735
    
    [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1145/1007730.1007735&link_type=DOI) 

50. [50].Harvey MA, Pierce M, Walter JE, et al. Obstetrical Anal Sphincter Injuries (OASIS): Prevention, Recognition, and Repair. Journal of Obstetrics and Gynaecology Canada. 2015;37(12):1131–1148. doi:10.1016/S1701-2163(16)30081-0
    
    [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1016/S1701-2163(16)30081-0&link_type=DOI) 
    
    [PubMed](http://medrxiv.org/lookup/external-ref?access_num=26637088&link_type=MED&atom=%2Fmedrxiv%2Fearly%2F2024%2F06%2F04%2F2024.06.03.24308379.atom) 

51. [51].Kropshofer S,  T Aigmüller, Beilecke K, et al. Management of Third and Fourth-Degree Perineal Tears After Vaginal Birth. Guideline of the DGGG, OEGGG and SGGG (S2k-Level, AWMF Registry No. 015/079, December 2020). Geburtshilfe Und Frauenheilkunde. 2022;83(02):165–183. doi:10.1055/a-1933-2647,
    
    [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1055/a-1933-2647&link_type=DOI) 

52. [52].Roper JC, Thakar R, Sultan AH. Under-classified obstetric anal sphincter injuries. Int Urogynecol J. 2022 Jun;33(6):1473–1479. doi: 10.1007/s00192-021-05051-y. Epub 2022 Feb 12
    
    [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1007/s00192-021-05051-y&link_type=DOI) 

53. [53].O’Leary BD, Kelly L, Fitzpatrick M, Keane DP.. Underdiagnosis of internal anal sphincter trauma following vaginal delivery.Ultrasound Obstet Gynecol. 2023 Feb;61(2):251–256. doi: 10.1002/uog.26049
    
    [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1002/uog.26049&link_type=DOI) 

54. [54].Kirss J, Pinta T, Böckelman C, Victorzon M. Factors predicting a failed primary repair of obstetric anal sphincter injury. Acta Obstet Gynecol Scand. 2016 Sep;95(9):1063–9. doi: 10.1111/aogs.12909. Epub 2016 May 18.
    
    [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1111/aogs.12909&link_type=DOI)