Reproducibility and reliability of automated volumetric measurement of single preovulatory follicles using SonoAVC

Transcription

1 Reproducibility and reliability of automated volumetric measurement of single preovulatory follicles using SonoAVC Samuel Salama, M.D., a Elisangela Arbo, M.D., a,b Frederic Lamazou, M.D., a Jean Marc Levailllant, M.D., a Rene Frydman, M.D., a and Renato Fanchin, M.D., Ph.D. a a AP-HP, Service de Gynecologie-Obstetrique et Medecine de la Reproduction, H^opital Antoine Beclere, Clamart, France; Univ Paris-Sud, Clamart, France; INSERM, Clamart, France; and b Programa de Pos-graduacxao em Ci^encias Medicas, Universidade Federal do Rio Grande do Sul, Porto Alegre, Brazil Objective: To evaluate the reproducibility and the reliability of an innovative, computer-assisted approach for automatically measuring ovarian follicles. Design: Prospective, comparative study. Setting: Hospital Beclere, Clamart, France. Patient(s): Fifteen infertile women undergoing IVF-ET in monodominant follicle cycles. Intervention(s): Just before oocyte retrieval, follicles were three-dimensionally reconstructed from transvaginal ultrasonographic images. Volumes were determined both manually by visual outlining of inner follicle borders (VOCAL) and automatically using SonoAVC. Each procedure was repeated three times. Follicular fluid volume indicated the actual follicle volume. Main Outcome Measure(s): Reproducibility and concordance of results were assessed by the intraclass correlation coefficient (ICC) and the limits of agreement method, respectively. Result(s): At any time, VOCAL (3.68, ml; 3.73, ml; 3.89, ml; median, ranges, respectively) and SonoAVC (3.57, ml; 3.71, ml; 4.07, ml, respectively) volume measurements failed to be statistically different from the corresponding actual follicle volume (3.60, ml). Reproducibility (ICC, 95% confidence intervals) of repeated VOCAL (0.95, ) and SonoAVC (0.97, ) measurements as well as 95% limits of agreement between actual volumes and VOCAL ( 0.48 to þ0.80 ml) and SonoAVC ( 0.61 to þ0.99 ml) measurements were comparable. Conclusion(s): Automatic measurement of ovarian follicle volumes from three-dimensionally reconstructed ultrasound images (SonoAVC) is a rapid and simple technique, which reproducibility and reliability are comparable to the semimanual technique (VOCAL). It opens new perspectives for the accurate and objective assessment of ovarian function by ultrasound. (Fertil Steril Ò 2010;93: Ó2010 by American Society for Reproductive Medicine.) Key Words: SonoAVC, VOCAL, 3D ultrasound, follicle volume, in vitro fertilization, preovulatory follicle Received September 28, 2008; revised December 22, 2008; accepted December 29, 2008; published online April 1, S.S. has nothing to disclose. E.A. has nothing to disclose. F.L. has nothing to disclose. J.M.L. has nothing to disclose. R.Fr. has nothign to disclose. R.Fa. has nothing to disclose. The present article was selected for oral presentation at the 64th annual meeting of the American Society for Reproductive Medicine, San Francisco, CA, November 8 12, Reprint requests: Renato Fanchin, M.D., Ph.D., Department of Obstetrics and Gynecology and Reproductive Medicine, H^opital Antoine Beclere, 157, rue de la Porte de Trivaux, 92141, Clamart, France (FAX: ; renato.fanchin@abc.aphp.fr). The objective, noninvasive appraisal of ovarian follicular status at ultrasound scans constitutes an invaluable parameter in the study of ovarian function. In assisted reproductive technologies (ART), it is a key element for determining follicle maturation and readiness for triggering ovulation (1). When performed on a routine basis, such a procedure has long been restricted to two-dimensional (2D), manual determination of follicle diameters, as follicle volume assessment remained unreliable, complex, and time-consuming. Yet, because ovarian follicles often display intricate shapes that depart from the ideal sphere or ellipsoid, manual determination of diameters only offers a pale reflection of follicle volume (2), a phenomenon that potentially misguides medical decisions. During the last decade, multiple refinements of transvaginal ultrasound transducers and signal-treatment softwares, in particular the Virtual Organ Computer-aided Analysis (VO- CAL) technology, have greatly improved both accuracy and user-friendliness of three-dimensional (3D) reconstruction of ovarian ultrasonographic images, thereby allowing the volumetric study of ovarian follicles (3, 4). Indeed, a previous study has demonstrated that VOCAL-assisted assessment of follicle volume provides more reliable results compared with those obtained from geometric extrapolation of 2D diameters (5). However, to determine follicle volume, and because of the often irregular shape of this anatomic structure, VOCAL technology requires the outlining of inner follicle borders, a manual process that has to be repeated from 6 to 30 times (every 30,15,9,or6 rotational angles) and that is, therefore, time-consuming and operator-dependent. This methodologic contingency prevents its routine use during the monitoring of assisted reproductive treatments /10/$36.00 Fertility and Sterility â Vol. 93, No. 6, April doi: /j.fertnstert Copyright ª2010 American Society for Reproductive Medicine, Published by Elsevier Inc.

2 To overcome these limitations, we tested an innovative approach derived from topologic segmentation technology that is complementary to VOCAL but allows, in a user-friendly interface, the automatic recognition of follicle borders, 3D follicle reconstruction, and follicle volume assessment (Sonography-based Automated Volume Count, SonoAVC). For this, we elected to evaluate the reproducibility of repeated SonoAVC results and to compare it with that of manual diameter and semimanual volume (VOCAL) measurements. In addition, we evaluated the concordance between actual follicle volumes and SonoAVC, manual diameter (2D), and semimanual volume (VOCAL) measurements, respectively. To avoid possible confounding effects of multiple preovulatory follicles on the reliability of our results, we preferred to use the experimental model of monodominant follicle in vitro fertilization-embryo transfer (IVF-ET), in which only one follicle reaches the preovulatory stage. MATERIALS AND METHODS Subjects Fifteen infertile women, 26 to 41 years of age, were studied prospectively. All of them met the following inclusion criteria: [1] both ovaries present, deprived of morphologic abnormalities, and adequately visualized in transvaginal ultrasound scans; [2] menstrual cycle length range between 25 and 35 days; [3] no current or past diseases affecting ovaries or gonadotropin or sex steroid secretion, clearance, or excretion; [4] no clinical signs of hyperandrogenism; [5] body mass indexes (BMI) ranging between 18 and 27.5 kg/m 2. An informed consent was obtained from all women, and this investigation received the approval of our internal institutional review board. Cycle Monitoring Cycle monitoring of monodominant IVF-ET cycles has been exhaustively described elsewhere (6). In brief, from cycle day 8 onward, selection of the dominant follicle was detected by transvaginal ultrasound. When the mean diameter of the dominant follicle exceeded 12 mm, to prevent the risk of premature LH peak and to control further follicular maturation, women were administered subcutaneously 0.5 mg of a GnRH antagonist (cetrorelix acetate; Cetrotide 0.25 mg, Serono Pharmaceuticals, Boulogne, France) and 150 IU of hmg (Menopur, Ferring Pharmaceuticals, Gentilly, France) daily until the day of hcg (Gonadotrophine Chorionique Endo, Organon Pharmaceuticals, Saint-Denis, France) administration. The choice of starting hmg treatment once follicle dominance had been achieved aimed at preventing the rescue of additional subordinated follicles. Eventually, women received a 5,000-IU hcg injection intramuscularly when the dominant follicle diameter ranged between 16 mm and 22 mm. The single oocyte was picked up approximately 34 hours after hcg administration and ET was performed 2 days after oocyte pickup. According to the present study s design, each patient had only one oocyte retrieved and only one embryo obtained and transferred. Luteal phase was supported with micronized progesterone (Estima Ge, Effik Pharmaceuticals, Bievres, France; 600 mg/day) administered daily by vaginal route starting on the evening of ET. Ultrasonographic Measurement of the Preovulatory Follicle Volume In the present investigation, by design, only one preovulatory follicle was obtained and measured in each patient. Within the 30 minutes before oocyte retrieval (approximately 9:00 a.m.), the preovulatory follicle was characterized using a MHz transvaginal ultrasound probe (RIC5-9H, General Electric Medical Systems, Paris, France) equipped with a 146 rotating head that allowed data acquisition for further 3D reconstruction. All examinations were performed by the same operator (S.S.). As depicted in Figure 1, for each follicle, three different types of measurements were made. First, we measured manually the mean follicular diameter (manual diameter measurement). For this, the two orthogonal diameters (d1 and d2) were determined at the largest follicle plane using 2D ultrasound by placing calipers at the inner follicle borders. Mean follicle diameter corresponded to (d1 þ d2)/2 and results were expressed in millimeters (mm). Second, preovulatory follicle volume was determined manually using offline 3D-reconstruction by means of a virtual organ computer aided analysis (VOCAL) technology (General Electric Healthcare, Kretz, Austria). For this, the region of interest in the 3D volume, which corresponded to the inner follicle border, was manually and progressively (every 30 ) set (VOCAL measurement). Third, the same follicle volume was reassessed automatically, in the same 3D-reconstructed images, using Sonography-based Automated Volume Count (SonoAVC) technology (General Electric Healthcare, Kretz, Austria), an approach that is based on topologic segmentation approach and that allows the automatic evaluation of follicle volumes (SonoAVC measurement). In brief, the SonoAVC algorithm identifies hypoechoic structures and their approximate shape (ovarian follicles) in the selected 3D matrix. Then, after having recognized the center of each structure, it is able to progressively calculate the exact number of surrounding voxels toward the structure borders (growth function) and extrapolate its volume. Volumes were expressed in milliliters (ml) and diameters were expressed in millimeters (mm). To assess reproducibility of measurements, we repeated each procedure (manual diameter, VOCAL, and SonoAVC measurements) three times for each preovulatory follicle. Estimation of the Actual Preovulatory Follicle Volume In the present investigation, actual preovulatory follicle volume was extrapolated from follicular fluid (FF) volume. Under transvaginal ultrasound guidance, FF from the single preovulatory follicle was thoroughly aspirated using a 10- ml syringe connected to a 35-mm, 17-gauge needle then maintained at steady temperature conditions (37 C) until 2070 Salama et al. Automatic measurement of follicles Vol. 93, No. 6, April 2010

3 FIGURE 1 The three types of measurements performed on the same preovulatory follicle: follicle 2D diameters (A), semimanual (VOCAL, B), and automatic (SonoAVC, C) volume measurements. Statistics The measure of central tendency used was the mean and the measure of variability was the standard error. Medians and ranges were used when normality of data distribution could not be ascertained. To assess the within-technique reproducibility for the three consecutive measurements, we calculated the intraclass correlation coefficient (ICC) (7) and its 95% confidence intervals (CI) for each parameter (manual diameter, VOCAL, and SonoAVC). The ICC is the ratio of the between-subject variability over the total variability, the latter including between- and within-subject variability, and it ranges from 0 to The ICC values exceeding 0.80 usually indicate adequate reproducibility. To assess concordance (reliability) between the actual preovulatory follicle volume (FF volume) with VOCAL and SonoAVC volumes, we used the limits of agreement method as reported by Bland and Altman (8) and ICC (95% CI) calculation. RESULTS Population Characteristics Patients were aged 36 (26 41) years, their BMI was 21.5 ( ) kg/m 2, and their infertility was because of male factor (n ¼ 7), tubal factor (n ¼ 3), endometriosis (n ¼ 1), dysovulation (n ¼ 3), and unexplained (n ¼ 1). Reproducibility of Measurements Reproducibility of the three consecutive manual diameter, semimanual volume (VOCAL), and automatic volume (SonoAVC) measurements of the preovulatory follicle, calculated by ICC and correspondent 95% CI, is presented in Table 1. Median manual diameters (ranges) were 18.4 ( ) mm, 18.9 ( ) mm, 19.4 ( ) mm, respectively, whereas VOCAL values were 3.68 ( ) ml, 3.73 ( ) ml, and 3.89 ( ) ml and SonoAVC values were 3.57 ( ) ml, 3.71 ( ) ml, and 4.07 ( ) ml, respectively. As shown in Table 1, manual diameter measurements corresponded to a lower ICC value (<0.80, which indicates poor reproducibility) compared with those of semimanual and automatic volume measurements. In addition, both VOCAL and SonoAVC volume measurements displayed adequate and comparable reproducibility with ICC values at 0.95 and 0.97, respectively. the oocyte was found and isolated. The FF volume was estimated by adding the volume contained in the aspiration syringe to the residual volume contained in the needle (0.5 ml). Reliability of Measurements The assessment of reliability of semimanual (VOCAL) and automatic (SonoAVC) volume measurements compared the actual volume of the preovulatory follicle, estimated by the Bland and Altman plot, is illustrated in Figure 2. At any time, the three VOCAL and SonoAVC volume results were not statistically different from actual follicular volumes (3.60, ml). Bland-Altman plot analysis indicated that the 95% limits of agreement between actual volumes and VOCAL ( 0.48 to þ0.80 ml) and SonoAVC ( 0.61 to þ0.99 ml) results were comparable (Table 2). The reliability Fertility and Sterility â 2071

4 TABLE 1 Reproducibility of the three consecutive manual diameter, manual volume (VOCAL), and automatic volume (SonoAVC) measurements of the preovulatory follicle. ICC 95% CI Manual diameters Manual volumes (VOCAL) Automatic volumes (SonoAVC) Note: ICC ¼ intraclass correlation coefficient; CI ¼ confidence interval. of VOCAL and SonoAVC volume measurement compared with the actual volume of the preovulatory follicle, estimated by the ICC calculation, were 0.96 ( ) and 0.95 ( ) respectively. DISCUSSION For long and for practical reasons, ovarian follicle diameters determined at 2D ultrasound scans remained the reference method to assess follicle sizes and, therefore, to anticipate the degree of follicle readiness to ovulation triggering. Yet, the correlation between the results of such a widespread technique in ART and the actual size of the preovulatory follicle volume is relatively poor (2, 5, 9). Indeed, preovulatory follicles are volumetric structures whose shape often is deformed by surrounding structures, such as other growing follicles as a result of controlled ovarian hyperstimulation (COH), cysts, pelvic organs, and so forth. Together with equipment features and operator skillfulness, the complex shapes frequently assumed by preovulatory follicles greatly contribute to explain the insufficient reliability of diametric measurements. Unfortunately, the first attempts to circumvent these limitations by determining the whole volume of ovarian follicles remained insufficient. Indeed, whereas the mathematic extrapolation of follicular volume from its 2D diameters is considered inaccurate (10), the use of 3D reconstruction technologies that require the freehand outlining of inner follicle borders (VOCAL) (9) are time-consuming and impractical for routine use. The present investigation evaluated both the reproducibility and reliability of an innovative, computer-assisted approach for automatically measuring the volume of preovulatory ovarian follicles (SonoAVC). We elected to study patients undergoing IVF-ET performed in single preovulatory follicle cycles rather than in COH cycles to make it simpler and more reliable to compare the results of different types of follicle measurements and the volume of follicular fluid aspirates. Our results indicated that automatic volume measurements (SonoAVC) are at least as reproducible and reliable as semimanual measurements (VOCAL), and both approaches showed a better reproducibility than routine 2D diameter measurements. These results confirm and expands those from a preliminary experience of the use of SonoAVC in the measurement of stimulated follicles of IVF-ET candidates (11). Moreover, the possible influence of other ultrasound machines, having other signal treatment specificities, on the reliability and reproducibility of SonoAVC constitute and important issue and should be addressed in further studies. The accurate and simple appraisal of the preovulatory follicle dimension using SonoAVC is by no means without consequences on the daily practice of ART. First, the manual, exhaustive recording of individual follicle diameters as it is practiced currently in the presence of the patient in most of ART centers worldwide may be time-consuming and imprecise. The possibility of rapidly scanning both ovaries for without-the-patient 3D image reconstruction and SonoAVC measurement of follicle volumes may constitute a practical option to improve reliability and friendliness of routine ovarian measurements. Further, because of the standardization of measurement technique and volumes information (12) FIGURE 2 Reliability of semimanual (VOCAL) and automatic (SonoAVC) volume measurements related to the actual volume (follicular fluid volume) of the preovulatory follicle, represented in Bland and Altman plots Salama et al. Automatic measurement of follicles Vol. 93, No. 6, April 2010

5 TABLE 2 Reliability of manual (VOCAL) and automatic (SonoAVC) volume measurements versus the actual volume of the preovulatory follicle, estimated by ICC calculation. Manual volume (VOCAL) actual volume measurements Automatic volumes (SonoAVC) actual volume measurements ICC 95% CI Note: ICC ¼ intraclass correlation coefficient; CI ¼ confidence interval. offered by SonoAVC, we can contemplate the possibility that intra- and interoperator variability could be reduced by this new approach. This issue deserves further investigation. Second, these technologic refinements instigate a complete revision of criteria used for determining follicle readiness for ovulation triggering, even if the relationship between follicle size and ART outcome has been the matter of debate. Whereas some investigators have observed a relationship between preovulatory follicle diameters (10, 13) or volumes (14) and oocyte/embryo competence in IVF-ET, in an extensive study including 9,933 follicles aspirated for IVF-ET, Salha et al. (15) were unable to relate follicle volume with oocyte preparedness to fertilization and embryo potential to implantation, despite the fact that in their series embryo implantation rates tended to be reduced when follicle volumes were either %1.0 or >5 ml. Additional prospective studies aiming at associating follicle volume measured by SonoAVC with oocyte/embryo competence are needed to verify the possible positive consequences of this methodology on the results of ART. In conclusion, our present results indicate that automatic measurement of ovarian follicle volumes from three-dimensionally reconstructed ultrasound images (SonoAVC) is a rapid and simple technique, which reproducibility and reliability are comparable to the semimanual technique (VO- CAL). Indeed, by extrapolation, in hyperstimulated ovaries, the possibility offered by SonoAVC of not having to measure manually the diameters of each follicle is very attractive (11). It opens new perspectives for the accurate and objective assessment of ovarian function by ultrasound, in particular, by encouraging the widespread use of follicle volumes instead of diameters in the monitoring of preovulatory follicle preparedness to controlled ovulation. These promising results suggest that simplicity and accuracy of SonoAVC will hopefully prompt clinicians to shift from 2D to 3D measurements in their daily assessment of such volumetric, irregular structures as ovarian follicles. Finally, the usefulness of SonoAVC in other clinical indications in reproductive medicine as the identification and counting of small antral follicles (16, 17), in an effort to rendering this important clinical indicator of ovarian reserve more reliable and operator-independent than conventional manual methodology, deserves further investigation. REFERENCES 1. Queenan JT, O Brien GD, Bains LM, Simpson J, Collins WP, Campbell S. Ultrasound scanning of ovaries to detect ovulation in women. Fertil Steril 1980;34: Penzias AS, Emmi AM, Dubey AK, Layman LC, DeCherney AH, Reindollar RH. Ultrasound prediction of follicle volume: is the mean diameter reflective? Fertil Steril 1994;62: Merce LT, Bau S, Barco MJ, Troyano J, Gay R, Sotos F, Villa A. Assessment of the ovarian volume, number and volume of follicles and ovarian vascularity by three-dimensional ultrasonography and power Doppler angiography on the hcg day to predict the outcome in IVF/ICSI cycles. Hum Reprod 2006;21: Jayaprakasan K, Hilwah N, Kendall NR, Hopkisson JF, Campbell BK, Johnson IR, et al. Does 3D ultrasound offer any advantage in the pretreatment assessment of ovarian reserve and prediction of outcome after assisted reproduction treatment? Hum Reprod 2007;22: Kyei-Mensah A, Zaidi J, Pittrof R, Shaker A, Campbell S, Tan SL. Transvaginal three-dimensional ultrasound: accuracy of follicular volume measurements. Fertil Steril 1996;65: Fanchin R, Mendez Lozano DH, Frydman N, Gougeon A, di Clemente N, Frydman R, et al. Anti-Mullerian hormone concentrations in the follicular fluid of the preovulatory follicle are predictive of the implantation potential of the ensuing embryo obtained by in vitro fertilization. J Clin Endocrinol Metab 2007;92: Shrout PE, Fleiss JL. Intraclass correlations: use in assessing rater reliability. Psychol Bull 1979;86: Bland JM, Altman DG. Statistical methods for assessing agreement between two methods of clinical measurement. Lancet 1986;1: Amer A, Hammadeh ME, Kolkailah M, Ghandour AA. Three-dimensional versus two-dimensional ultrasound measurement of follicular volume: are they comparable? Arch Gynecol Obstet 2003;268: Wittmaack FM, Kreger DO, Blasco L, Tureck RW, Mastroianni L Jr, Lessey BA. Effect of follicular size on oocyte retrieval, fertilization, cleavage, and embryo quality in in vitro fertilization cycles: a 6-year data collection. Fertil Steril 1994;62: Raine-Fenning N, Jayaprakasan K, Clewes J, Joergner I, Bonaki SD, Chamberlain S, et al. SonoAVC: a novel method of automatic volume calculation. Ultrasound Obstet Gynecol 2008;31: Raine-Fenning NJ, Campbell BK, Clewes JS, Johnson IR. The interobserver reliability of ovarian volume measurement is improved with three-dimensional ultrasound, but dependent upon technique. Ultrasound Med Biol 2003;29: Dubey AK, Wang HA, Duffy P, Penzias AS. The correlation between follicular measurements, oocyte morphology, and fertilization rates in an in vitro fertilization program. Fertil Steril 1995;64: Ectors FJ, Vanderzwalmen P, Van hoeck J, Nijs M, Verhaegen G, Delvigne A, et al. Relationship of human follicular diameter with oocyte fertilization and development after in-vitro fertilization or intracytoplasmatic sperm injection. Hum Reprod 1997;12: Salha O, Nugent D, Dada T, Kaufmann S, Levett S, Jenner L, et al. The relationship between follicular fluid aspirate and oocyte maturity in in-vitro fertilization cycles. Hum Reprod 1998;13: Reuss ML, Kline J, Santos R, Levin B, Timor-Tritsch I. Age and the ovarian follicle pool assessed with transvaginal ultrasonography. Am J Obstet Gynecol 1996;174: Jayaprakasan K, Campbell BK, Clewes JS, Johnson IR, Raine- Fenning NJ. Three-dimensional ultrasound improves the interobserver reliability of antral follicle counts and facilitates increased clinical work flow. Ultrasound Obstet Gynecol 2008;31: Fertility and Sterility â 2073