Differential diagnosis of middle compartment pelvic organ prolapse with transperineal ultrasound

Artificial Intelligence to Determine Correct Midsagittal Plane in Dynamic Transperineal Ultrasound

PURPOSE

To create and validate a machine learning(ML) model that allows for identifying the correct capture of the midsagittal plane in a dynamic ultrasound study, as well as establishing its concordance with a senior explorer and a junior explorer.

METHODS

Observational and prospective study with 90 patients without pelvic floor pathology. Each patient was given an ultrasound video where the midsagittal plane of the pelvic floor was recorded at rest and during the Valsalva maneuver. A segmentation model was used that was trained on a previously published article, generating the segmentations of the 90 new videos to create the model. The algorithm selected to build the model in this project was XGBoost(Gradient Boosting). To obtain a tabular dataset on which to train the model, feature engineering was carried out on the raw segmentation data. The concordance of the model, of a junior examiner and a senior examiner, with the expert examiner was studied using the kappa index.

RESULTS

The first 60 videos were used to train the model and the last 30 videos were reserved for the test set. The model presented a kappa index 0.930(p < 0.001) with very good agreement for detection of the correct midsagittal plane. The junior explorer presented a very good agreement (kappa index = 0.930(p < 0.001)). The senior explorer presented a kappa index 0.789(p < 0.001) (good agreement) for detection of the correct midsagittal plane.

CONCLUSION

We have developed a model that allows determining the correct midsagittal plane captured through dynamic transperineal ultrasound with a level of agreement comparable to or greater than that of a junior or senior examiner, using expert examiner assessment as the gold standard.

Applicability of Deep Learning to Dynamically Identify the Different Organs of the Pelvic Floor in the Midsagittal Plane

INTRODUCTION AND HYPOTHESIS

The objective was to create and validate the usefulness of a convolutional neural network (CNN) for identifying different organs of the pelvic floor in the midsagittal plane via dynamic ultrasound.

METHODS

This observational and prospective study included 110 patients. Transperineal ultrasound scans were performed by an expert sonographer of the pelvic floor. A video of each patient was made that captured the midsagittal plane of the pelvic floor at rest and the change in the pelvic structures during the Valsalva maneuver. After saving the captured videos, we manually labeled the different organs in each video. Three different architectures were tested-UNet, FPN, and LinkNet-to determine which CNN model best recognized anatomical structures. The best model was trained with the 86 cases for the number of epochs determined by the stop criterion via cross-validation. The Dice Similarity Index (DSI) was used for CNN validation.

RESULTS

Eighty-six patients were included to train the CNN and 24 to test the CNN. After applying the trained CNN to the 24 test videos, we did not observe any failed segmentation. In fact, we obtained a DSI of 0.79 (95% CI: 0.73 - 0.82) as the median of the 24 test videos. When we studied the organs independently, we observed differences in the DSI of each organ. The poorest DSIs were obtained in the bladder (0.71 [95% CI: 0.70 - 0.73]) and uterus (0.70 [95% CI: 0.68 - 0.74]), whereas the highest DSIs were obtained in the anus (0.81 [95% CI: 0.80 - 0.86]) and levator ani muscle (0.83 [95% CI: 0.82 - 0.83]).

CONCLUSIONS

Our results show that it is possible to apply deep learning using a trained CNN to identify different pelvic floor organs in the midsagittal plane via dynamic ultrasound.

Use of transrectal high-frequency ultrasound in assessment of middle compartment prolapse

OBJECTIVE

To investigate the superiority of transrectal high-frequency ultrasound (TRUS) in precise assessment of middle compartment prolapse in comparison with routine transperineal ultrasound (TPUS).

METHODS

Prospectively analyzed and compared detection rates of entire cervical length and uterine descent on TPUS and TRUS in 101 patients with pelvic organ prolapse (POP).

RESULTS

Detection rates of entire cervix on TRUS were significantly higher than those on TPUS both at rest and during Valsalva maneuver (90.10% VS 49.50%, 92.08% VS 9.90% respectively, both p < 0.05). Uterine descent was able to be evaluated in 92.08% of patients by TRUS and in 5.94% of patients by TPUS, which was statistically significant (p < 0.05). The interobserver repeatability for the measurements of anterior lip, cervical canal and posterior lip on TRUS was excellent. The mean lengths of anterior lip, cervical canal and posterior lip were significantly increased during Valsalva maneuver than those measured at rest (p < 0.05). And mean length of anterior lip was longer than posterior lip both at rest and during Valsalva (p < 0.05).

CONCLUSION

TRUS can significantly raise detection rates of entire cervix, and make the direct evaluation of uterine descent feasible. TRUS can be used as a complementary method to TPUS to attain more comprehensive and accurate presurgical imaging information in middle compartment prolapse patients.

Is the Presence of Levator Ani Muscle Avulsion Relevant for the Diagnosis of Uterine Prolapse?

OBJECTIVE

To determine if the addition of the assessment of levator ani muscle (LAM) avulsion to the measurement of the difference in the pubis-uterine fundus distance between rest and with the Valsalva maneuver could increase the diagnostic capacity of ultrasound for uterine prolapse (UP).

METHODS

This multicenter, observational and prospective study included 145 patients. Ultrasound assessment was performed, establishing the diagnosis of UP as a difference between the pubic-uterine fundus distance at rest and during the Valsalva maneuver ≥15 mm (standard technique), while LAM avulsion was defined as an abnormal LAM insertion in three central slices using multislice ultrasound. A binary multivariate logistic regression model was made using nonautomated methods to predict surgical UP (general population, premenopausal, and postmenopausal patients), including the difference between the pubis-uterine fundus distance at rest and with the Valsalva maneuver as well as LAM avulsion.

RESULTS

A total of 143 patients completed the study. The addition of LAM avulsion criteria to the standard dynamic distance-based protocol for the diagnosis of UP resulted in a higher sensitivity for the general population (79.7 vs 68.1%) as well as for premenopausal (89.3 vs 79.9%) and postmenopausal patients (76 vs 66.1%). In contrast, the standard technique showed a higher specificity than the model based on the standard technique associated with LAM avulsion for the general population (89.2 vs 74.3%) and premenopausal women (91.7 vs 63.2%). For postmenopausal patients, the model based on the standard technique associated with LAM avulsion had a higher sensitivity (76 vs 66.1%) and specificity (91.7 vs 86.8%) than the ultrasound diagnosis of UP.

CONCLUSION

The implementation of the assessment of LAM avulsion in the ultrasound diagnosis of UP is useful in postmenopausal patients, increasing sensitivity and specificity relative to the ultrasound assessment based only on the difference between the pubis-uterine fundus distance at rest and with the Valsalva maneuver.

Is It Possible to Diagnose Surgical Uterine Prolapse With Transperineal Ultrasound? Multicenter Validation of Diagnostic Software

OBJECTIVES

To validate an ultrasound software that uses transperineal ultrasound to diagnose uterine prolapse (UP).

METHODS

Multicenter, observational and prospective study with 155 patients that had indications for surgical intervention for dysfunctional pelvic floor pathology. Each patient underwent an examination with Pozzi tenaculum forceps was performed in the operating room with the patient anesthetized, followed by surgical correction of stages II-IV UP. Transperineal ultrasound was used to assess the difference in the pubis-uterine fundus measurement. With a multivariate logistic regression binary model (with the measurement ultrasound at rest, the Valsalva maneuver and age) using nonautomated methods to predict UP. With the purpose of evaluating the model, a table with coordinates of the receiver operating characteristic (ROC) curve, after which sensitivity and specificity were assessed.

RESULTS

A total of 153 patients were included (73 with a diagnosis of surgical UP). It was obtained from the AUC (0.89) of the probabilities predicted by the model (95% confidence interval, 0.84-0.95; P < .0005). Based on the ROC curve for the model, obtaining a sensitivity of 91.8% and a specificity of 72.7%, values that were superior to those for the clinical exam for surgical UP (sensitivity: 80.8%; specificity: 71.3%).

CONCLUSIONS

We validated software that uses transperineal ultrasound of the pelvic floor and patient age to generate a more reliable diagnosis of surgical UP than that obtained from clinical examinations.

Identification of Surgical Uterine Prolapse in Premenopausal Patients With Clinical or Ultrasound Criteria? A Multicenter Comparative Study

OBJECTIVES

It is unknown whether diagnosing uterine prolapse (UP) via ultrasound or surgical criteria is superior. Our objective is to determine whether the diagnostic capacity of ultrasound with surgical criteria differs from that of surgical criteria only.

METHODS

This was a multicenter prospective observational study with 54 premenopausal patients with surgical criteria for a dysfunctional pelvic floor pathology who were consecutively recruited for 1 year. Clinical UP with surgical criteria was defined when UP stage II-IV was identified (during pelvic floor consultation), and UP diagnosed by ultrasound with surgical criteria was established when a difference ≥15 mm was found between rest and Valsalva applied to the pubis-uterine fundus. The sensitivity, specificity and positive and negative predictive values were determined to evaluate clinical and ultrasound methodologies as diagnostic tests.

RESULTS

UP diagnosed by ultrasound with surgical criteria presented better sensitivity (78.57 vs 35.71%), specificity (92.11 vs 81.58%), positive predictive value (61.83 vs 23.99%), and negative predictive value (96.35 vs 11.37%) than UP diagnosed by surgical criteria only.

CONCLUSION

Ultrasound with surgical criteria is superior to surgical criteria alone when diagnosing UP.

Predictive Model for the Diagnosis of Uterine Prolapse Based on Transperineal Ultrasound

We want to describe a model that allows the use of transperineal ultrasound to define the probability of experiencing uterine prolapse (UP). This was a prospective observational study involving 107 patients with UP or cervical elongation (CE) without UP. The ultrasound study was performed using transperineal ultrasound and evaluated the differences in the pubis−uterine fundus distance at rest and with the Valsalva maneuver. We generated different multivariate binary logistic regression models using nonautomated methods to predict UP, including the difference in the pubis−uterine fundus distance at rest and with the Valsalva maneuver. The parameters were added progressively according to their simplicity of use and their predictive capacity for identifying UP. We used two binary logistic regression models to predict UP. Model 1 was based on the difference in the pubis−uterine fundus distance at rest and with the Valsalva maneuver and the age of the patient [AUC: 0.967 (95% CI, 0.939−0.995; p < 0.0005)]. Model 2 used the difference in the pubis−uterine fundus distance at rest and with the Valsalva maneuver, age, avulsion and ballooning (AUC: 0.971 (95% CI, 0.945−0.997; p < 0.0005)). In conclusion, the model based on the difference in the pubis−uterine fundus distance at rest and with the Valsalva maneuver and the age of the patient could predict 96.7% of patients with UP.

2D ultrasound diagnosis of middle compartment prolapse: a multicenter study

BACKGROUND

Recently, a specific methodology has been defined, using transperineal ultrasound, for the differential diagnosis of middle compartment prolapse [uterine prolapse (UP) or cervical elongation (CE) without UP] based on the difference in the pubis-uterine fundus distance at rest and with the Valsalva maneuver, with a cutoff point of 15 mm. The objective of this study was to validate the diagnostic utility of a ≥15 mm difference between the pubis-uterine fundus distance at rest and during the Valsalva maneuver to define UP in a multicenter study.

METHODS

This prospective multicenter observational study included 94 patients (UP =51; CE without UP =43). The clinical examination was based on the International Continence Society Pelvic Organ Prolapse Quantification (ICS POP-Q) system for assessing pelvic organ prolapse (POP) and patients were candidates for corrective surgery of the middle compartment of the pelvic floor (correction of UP or CE without UP). The ultrasound study was performed by transperineal ultrasound (B-mode) with the patient undergoing dorsal lithotomy. The distance evaluation was performed in relation to the posteroinferior pubic margin in the midsagittal plane, with reference to the uterine fundus (established as the most distal hyperechogenic) line from the pubis to the uterine fundus at rest and with the Valsalva maneuver. We defined UP detected using UP as a difference of ≥15 mm between the pubis-uterine fundus distance at rest and with the Valsalva maneuver. Agreement between the clinical and ultrasound diagnosis of UP was assessed using the Cohen kappa coefficient of agreement and its 95% CIs.

RESULTS

The ultrasound diagnosis of global UP at the three centers showed very good agreement, with a kappa index of 0.826 (0.71, 0.94). The agreement of ultrasound with the clinical diagnosis of UP using the ICS POP-Q system was very good for each of the hospitals [Hospital 1: 0.814 (0.64, 0.98), Hospital 2: 0.847 (0.64, 1) and Hospital 3: 0.824 (0.59, 1)].

CONCLUSIONS

A difference of ≥15 mm between the pubis-uterine fundus distance at rest and during the Valsalva maneuver for the diagnosis of UP presents very good agreement with the results of clinical evaluation with the ICS POP-Q system.

Interobserver variability of ultrasound measurements for the differential diagnosis of uterine prolapse and cervical elongation without uterine prolapse

OBJECTIVES

Our study aims to determine the interobserver variability of different ultrasound measurements (pubis-cervix distance, pubis-uterine fundus distance, and pubis-Douglascul-de-sac distance) previously analyzed for the ultrasound differential diagnosis of uterine prolapse (UP) and cervical elongation CE without UP.

MATERIALS AND METHODS

We conducted a prospective observational study with 40 patients scheduled to undergo surgical correction of UP and CE without UP. All patients underwent pelvic floor ultrasound examination by an examiner (E1) who acquired ultrasound images. Using these images, E1 measured the distances for the ultrasound differential diagnosis of UP and CE without UP, and these distances were compared with those measured by the other examiner (E2). Values were analyzed by calculating ICCs with 95% CIs.

RESULTS

For UP, excellent reliability was obtained for all measurements except the pubis-Douglascul-de-sac measurement at rest, which was moderate (ICC 0.596; p = 0.028) and for the difference between the pubis-Douglascul-de-sac measurement at rest and during the Valsalva maneuver, which was good (ICC 0.691; p < 0.0005). For CE without UP, interobserver reliability was excellent for all measurements analyzed except the pubis-cervix measurement during the Valsalva maneuver, which was moderate (ICC 0.535; p = 0.052) and for the pubis-Douglascul-de-sac measurement at rest, which was good (ICC 0.768; p < 0.0005).

CONCLUSIONS

There is excellent interobserver reliability in measurements of the difference in the distance from the pubic symphysis to the uterine fundus at rest and during the Valsalva maneuver for both UP and CE without UP, which are used for the ultrasound differential diagnosis of UP and CE without UP.