Congenital uterine anomalies are associated with an increased risk of repeated first and second trimester miscarriage and preterm delivery 19. A meta-analysis of published retrospective data comparing pregnancy outcome before and after hysteroscopic septoplasty indicated a marked improvement after surgery, which itself has minimal postoperative sequelae 20. Accurate and reliable diagnosis is important therefore as it allows the identification of patients at risk of these complications and timely surgical intervention. This is undoubtedly the area where three-dimensional sonography has contributed the most and has become the investigation of choice in units where available. This reflects its ability to demonstrate both the endometrial cavity and the myometrium simultaneously in the coronal plane as shown by Jurkovic et al. in their pioneering study of 61 patients with a history of recurrent miscarriage or infertility (Figs. 7) 21. Hysterosalpingography had shown the presence of a normal uterus in 44 (72.1%) patients, an arcuate uterus in nine (14.8%) and a major fusion defect in three cases (4.9%). Whilst two-dimensional sonography was associated with five false-positive diagnoses of arcuate uteri and three of major uterine anomalies, three-dimensional sonography agreed with hysterosalpingography in all of these cases. Shortly afterwards Raga et al. and Wu et al. both reported similarly favourable results suggesting three-dimensional sonography offered a 100% specificity for the exclusion of uterine anomalies and was able to differentiate between the different anomalies 2223.

Three-dimensional sonography has since been used to determine the prevalence of uterine anomalies in various patient groups and to characterise outcome on the basis of the anomaly. As many as 24% of women with recurrent pregnancy loss may have uterine anomalies 14 which is roughly four times that seen in low-risk women where the prevalence is in the order of 5 to 6% 2526. In terms of the type of anomaly a similar distribution is seen between different groups with arcuate uteri being the most common, followed by subseptate then bicornuate uteri with the more complex anomalies such as uterus didelphys and single uterine horns the least prevalent. Women with a subseptate uterus have a significantly higher proportion of first-trimester loss and women with an arcuate uterus a significantly greater proportion of second-trimester loss (p < 0.01) and preterm labor (p < 0.01) compared to women with a normal uterus 24. Another important finding derived from these three-dimensional studies has been that outcome is related not only to the degree of defect but also to the remaining cavity length which is significantly shorter in both arcuate and subseptate uteri in women with recurrent miscarriage 26. These measurement techniques and classifications used allow comparison between these studies as a degree of standardisation has been used, based on measurement of fundal distortion from the mid-point of an imaginary horizontal line joining the upper aspects of the cornuae to the upper aspect of the uterine cavity, that has been shown to be reliable between observers examining stored three-dimensional datasets (Fig. 7f) 27.

If we revisit the advantages proposed at the start of the review we can see that the majority are already satisfied in respect to uterine anomalies. Three-dimensional sonography offers a reliable and standardised tool to diagnose, differentiate and quantify uterine anomalies. Three-dimensional sonography has significantly added to our understanding of uterine anomalies qualifying their effect on reproductive outcome and thereby helping the clinician counsel patients accordingly and confidently.

La Torre et al. compared three-dimensional sonography with conventional imaging with and without saline contrast in their study of twenty-three patients in whom subsequent hysteroscopy revealed the presence of 16 endometrial polyps 28. Standard two-dimensional sonography demonstrated a relatively poor specificity of only 69.5% suggesting the presence of polyps in 23 patients. This was improved to 94.1% when two-dimensional sonography was used in conjunction with saline infusion as only 17 patients were then thought to have polyps. Three-dimensional sonography performed almost as well diagnosing the presence of polyps in 18 patients with a specificity of 88.8% and subsequently correctly identified all 16 polyps when used in conjunction with saline infusion. A similar improvement in specificity with three-dimensional sonography has been shown by Sylvestre et al. in their study of 209 subfertile patients thought to have an intrauterine lesion on transvaginal two-dimensional sonography or hysterosalpingography 29. Using saline infusion sonography with two-dimensional and then three-dimensional sonography, 92 patients were subsequently identified as having a variety of intrauterine lesions suggesting a sensitivity and specificity of 97% and 11% for two-dimensional sonography, 87% and 45% for three-dimensional sonography and 98% and 100% for two-dimensional saline infusion sonography. Of 59 patients that had undergone hysteroscopy, the sensitivity and positive predictive value of saline infusion sonography were 98% and 95% when performed in combination with two-dimensional sonography and 100% and 92% with three-dimensional sonography respectively. The study clearly demonstrates how simple contrast media potentially increase the specificity of two-dimensional sonography as 55% (116 of 209) of patients were found to have normal cavities following the infusion of saline. This was largely due to the correct localization of leiomyomas as intramural rather than submucosal (54 of 101). This is extremely important in the clinical setting, as whilst three-dimensional sonography improves further on two-dimensional imaging it is not currently available in all units and its use therefore has significant implications in terms of cost and training. Saline infusion on the other hand may be undertaken easily using various dedicated systems or modifications of readily available cheaper catheters such as a paediatric feeding tube. Figure 8 shows the typical three-dimensional sonography appearances of endometrial polyps at different stages of the menstrual cycle whilst Figure 9 demonstrates a sub-mucous fibroid at three-dimensional saline infusion sonography.

The combination of contrast media and three-dimensional sonography has also been used to assess tubal patency. Kiyokawa et al. found three-dimensional saline sonohysterosalpingography was able to demonstrate the entire contour of the uterine cavity in 96% of cases compared to only 64% cases with conventional X-ray hysterosalpingography (p < 0.005) and was associated with a positive predictive value and specificity of predicting tubal patency of 100% in 25 unselected infertile patients 30. Sladkevicius et al. found three-dimensional power Doppler imaging demonstrated free spill almost twice as often as conventional imaging (114 versus 58 tubes respectively) when used during hysterosalpingo-contrast sonography 31. Sankpal et al. reported less promising results with the same technique when they reassessed tubal patency in 15 women who had normal X-ray HSG examinations within the previous year 32. An important distinction was their use of saline rather than a positive contrast agent but it maybe that the technique has a distinct learning curve and numbers were small.

Several groups have used three-dimensional sonography to demonstrate ovarian volume and vascularity are increased in polycystic ovarian syndrome 3336. Three-dimensional sonography also allows for the measurement of stromal volume through the calculation and subtraction of total follicular volume from total ovarian volume (Fig. 10). Using this technique Kyei-Mensah et al showed stromal volume was positively correlated with serum androstenedione concentrations (p < 0.01) in 26 women with clinical evidence of polycystic ovaries 37. However, using a similar approach, Nardo et al. were unable to demonstrate any relationship between serum FSH, LH or testosterone and ovarian stromal volume in 23 infertile women with clomiphene citrate-resistant PCOS at the same stage of the menstrual cycle 38.

The first stage is to obtain control over the hypothalamic-pituitary-ovarian axis through the induction of pituitary 'down-regulation' achieved through a continuous and highly supraphysiological dose of a gonadotrophin releasing hormone (GnRH) analogue. Follicular development is then induced through the administration of natural or synthetic Follicle Stimulating Hormone (FSH) until at least three large follicles measuring 18 mm or more are present when the resumption of meiosis is initiated though a single dose of human Chorionic Gonadotrophin (hCG), which has a similar structure and pharmacodynamic effect to Luteinising Hormone (LH) but is more readily available. Exposure to hCG is essential prior to the collection of the oocytes to ensure that they are mature and therefore capable of being fertilised when incubated with sperm during in vitro fertilisation (IVF) or when injected with a single sperm in intracytoplasmic sperm injection (ICSI) treatment. This would eventually lead to ovulation in vivo if the oocytes were not physically retrieved and this procedure is critically timed therefore to occur within 34 to 36 hours of the administration of hCG.

The dose and duration of FSH treatment is dependent upon each individual patients response to stimulation and is adjusted dynamically during treatment dependent on the number of follicles developing and on the amount of oestrogen they produce collectively as measured in the serum. The response and eventual number of mature oocytes retrieved is referred to as the patients 'ovarian reserve'. Prediction of ovarian reserve has obvious importance as it allows treatment to be tailored to the individual potentially increasing the number of oocytes retrieved.

Fertilisation of an oocyte is confirmed by the presence of two pronuclei 18 to 24 hours after IVF or ICSI. The embryos are allowed to develop in vitro until the second or third day after oocyte retrieval when most units would transfer a maximum of two to the recipient uterus selected on the basis of their quality as determined by a subjective grading system. The chance of the implantation occurring is not only dependent upon the embryo stage and grade but also on the endometrium. Just how important 'endometrial receptivity' is in the overall process remains uncertain but the endometrium is likely to play more than a passive role.

With a better understanding of processes involved in controlled ovarian stimulation and assisted reproduction let us now turn to the application of three-dimensional sonography and critically review its current role.

Of the sonographic markers suggested as predictive of ovarian response the three that have been specifically addressed by three-dimensional sonographic studies are antral follicle counts, ovarian volume and ovarian blood flow.

Pellicer et al. were amongst the first to use three-dimensional sonography as an adjunct to conventional markers of ovarian reserve when they examined ovarian volume and the number of 'selectable follicles' measuring 2–5 mm in a small group of low responders on day three of the menstrual cycle 39. Both the number of selectable follicles and the total number of antral follicles were significantly decreased in the 'low responder' group who also demonstrated significantly higher serum FSH levels despite having values within the normal range. Ovarian volume measurements, however, were similar between the two groups. Pohl et al. also used three-dimensional sonography to quantify the number of follicles of varying diameter in 113 patients following 'down-regulation' but prior to ovarian stimulation (40. Patients with a higher number of follicles measuring between 5 and 10 mm were younger (p < 0.01), had a significantly higher number of oocytes retrieved (p < 0.0001) and were more likely to conceive (p < 0.05) (Fig. 12). Kupesic et al also suggest that the antral follicle count is a better predictor than three-dimensional measures of ovarian volume and blood flow 41. A minimal ovarian volume may be important however (Fig. 11). Schild et al. noted a pregnancy rate of only 6.7% (1 of 15) in patients with a minimum unilateral ovarian volume of ≤ 3 cm³, which represented a single standard deviation below the mean, versus 21.9% (30 of 137) in patients with an initial minimum ovarian volume above 3 cm³42. This difference was not significant however and cancellation rates due to poor ovarian response or failed fertilisation were similar in both groups.

There is no doubt that antral follicle counts, when used in categorical classifications, are an important predictor of 'ovarian reserve' and may be measured with a high level of agreement both between and within observers 43. Tree-dimensional sonography, however, does not appear to offer any significant advantage over two-dimensional imaging even at higher follicle counts when interobserver reliability is reduced. Ovarian volume has a limited predictive ability that does not appear to supersede that of antral follicle counts. Do measurements of ovarian vascularity add anything to validate the use of three-dimensional sonography as a marker of ovarian reserve?

Jarvela et al. used three-dimensional power Doppler angiography after pituitary 'down-regulation' and during gonadotrophin stimulation to compare ovarian vascularity in 33 women with normal ovarian reserve, as judged by antral follicle counts, to 12 women who had demonstrated a previous poor response 44. The number of oocytes retrieved correlated with the antral follicle count (R = 0.458, p < 0.01) and ovarian volume (R = 0.388, p < 0.05) but not with ovarian vascularity. All three indices of vascularity were shown to increase significantly during gonadotrophin stimulation in the group with normal ovarian reserve only but this was related to the antral follicle count reiterating the importance of this marker as an independent variable. Kupesic et al. similarly showed the number of oocytes retrieved and subsequent conception rate to be greater in patients with a greater ovarian volume and a greater ovarian stromal vascularity but not independently of a higher number of antral follicles 45.

Schild et al. measured endometrial thickness and volume in a total of 47 IVF cycles on the day of oocyte retrieval 46. There were no significant differences between the group of fifteen patients that conceived (31.9%) and the remaining 32 non-pregnant women in terms of the mean endometrial thickness (10.8 ± 2.3 mm versus 11.8 ± 3.4 mm) or volume (4.9 ± 2.2 cm³ versus 5.8 ± 3.4 cm³) respectively. Yaman et al. reported similar findings with no differences in endometrium volume (4.16 ± 1.97 cm³ versus 4.53 ± 1.79 cm³), endometrium thickness (11 ± 2 mm versus 11 ± 2 mm) between 21 pregnant and 44 non-pregnant women on the day of hCG administration 47. However, whilst there was no absolute endometrial thickness required for pregnancy a minimal endometrial volume above 2.5 cm³ favoured pregnancy. Raga et al. also described a cut-off point in endometrial volume in their study of 72 patients on the day of embryo transfer 48. Implantation rates were significantly lower (p < 0.05) in those women with endometria measuring less than 2 cm³ and no pregnancies were observed at volumes below 1 cm³. Even if there is a lower limit in endometrial volume below which pregnancy will not occur conventional measurement of endometrial thickness is already known to have a similar negative predictive value with conception less likely in patients with an endometrium measuring less than 5 mm in diameter.

Kupesic et al. reported more predictive information could be derived at the time of embryo transfer when three-dimensional power Doppler was used to quantify endometrial vascularity 49. Of 89 patients studied successful conception cycles were associated with a significantly higher endometrial flow index (13.2 ± 2.2 versus 11.9 ± 2.4, p < 0.05). Wu et al. also found three-dimensional power Doppler angiography to be an important determinant of 'endometrial receptivity' but on the day of hCG administration in 54 patients undergoing their first IVF cycle 50. The subendometrial vascularisation flow index (VFI) proved the best predictor of conception being superior to the vascularisation index (VI), flow index (FI) and endometrial volume in the receiver operating characteristics curve analysis. Interestingly, three-dimensional sonography may also be used to examine endometrial vascularity and determine 'endometrial receptivity' prior to ovarian stimulation. Schild et al. reported significantly lower indices of vascularity at down-regulation in 15 patients that subsequently conceived (20%) than in 60 non-conception cycles (p < 0.05) with the flow index the strongest predictive factor of IVF success (p < 0.05) 51. Endometrial measurements were once again not correlated with outcome. This may reflect a more profound pituitary suppression but is more likely to reflect patients responsive to exogenous hormonal therapy.

Gestational sac volume has been proposed as representative of the competence of the early uteroplacental unit and therefore a potential predictor of pregnancy outcome 52. In one of the first reported applications of three-dimensional sonography, Steiner et al. studied 38 pregnancies between 5 and 11 weeks gestation and found gestational sac volume measurements to significantly correlated with gestational age (r = 0.74, p < 0.001) and more than two standard deviations below the mean in three of five embryonic or anembryonic pregnancy failures 52. However, Acharya and Morgan have recently shown a clear relationship between gestational sac volume and sac diameter also exists in patients with embryonic pregnancy failure and that measurements are unable to predict eventual outcome during conservative management 53. Babinszki et al. did find a relationship between gestational sac and yolk sac volume and adverse outcome in a group of 49 patients who had conceived following treatment for subfertility but crown rump length performed equally as well 54.

Currently, therefore, one may conclude that whilst there is a distinct relationship between gestational age and three-dimensional measurements of gestational and yolk sac volume, these parameters do not appear to improve upon the predictive value of current tests in determining the eventual outcome in either viable or non-viable intrauterine pregnancies.

The improved spatial orientation afforded by three-dimensional sonography has been used to evaluate endometrial shape in cases of pregnancy of unknown location in an attempt to improve the sensitivity and specificity of sonography. Rempen et al. described the persistence of a distinct symmetry with regard to the median longitudinal axis of the uterus in the coronal plane in 90% of extrauterine pregnancies which was lost in intrauterine pregnancies 55. Su et al. reported a case of an anembryonic cervical pregnancy diagnosed at 10 weeks associated with a large arteriovenous malformation managed conservatively with selective uterine artery embolisation 56. Three-dimensional power Doppler ultrasonography proved useful in the initial diagnosis subsequent monitoring of the response to treatment. These studies suggest a potential role for three-dimensional sonography but in the absence of properly blinded data it is impossible to predict whether it will improve upon conventional sonography. The spatial orientation and additional information afforded by the coronal plane do lend themselves to assessment of the endometrial cavity as illustrated so elegantly by the three-dimensional studies on uterine anomalies so one should be optimistic in this regard. Certainly in our own unit we have found three-dimensional imaging to be beneficial in several cases of pregnancy of unknown location and particularly so in twin pregnancies following assisted reproduction treatment and embryo transfer where heterotropic pregnancy is so much more likely (Fig. 13).