Background

Polycystic ovarian syndrome [PCO] is a global complex endocrine disorder in women of reproductive age [1]. The hallmarks of the disease include menstrual disturbances, multiple ovarian cysts, and excess androgen. Women suffer from acne, hirsutism, the inability to conceive, and obesity. They also have high serum levels of free testosterone [T], androstenedione, dehydroepiandrosterone sulfate [DHEAS], luteinizing hormone [LH], and the LH/FSH ratio. The high serum level of anti-Mullerian hormone [AMH] is used as a marker for the increased total number of ovarian follicles, thus reflecting the severity of PCO. The metabolic aspects of PCO include obesity, insulin resistance, and diabetes mellitus type II. In addition, there is an increased risk of endometrial hyperplasia, cardiovascular diseases, and cerebrovascular disorders [2,3,4,5,6]. According to the presence of oligo-anovulation [O], clinical or biochemical signs of hyperandrogenism [HA], and polycystic ovarian morphology [PCOM], the National Institute of Health [NIH] in 2012 classified PCO into four phenotypes (A, B, C, and D). Phenotype A includes [O + HA + PCOM], phenotype B includes [O + HA], phenotype C includes [HA + PCOM], and phenotype D includes [O + PCOM]. The classic PCO [phenotype A/B] is manifested by more obvious menstrual dysfunction, hyperandrogenism, an abnormal lipid profile, and an increased body mass index. Furthermore, there is an elevated level of insulin and anti-Mullerian hormones, with an increased risk of other complications. However, the non-classic PCO is associated with less endocrine and metabolic dysfunction and a lower incidence of complications [7,8,9,10,11,12].

Hyperandrogenism has an essential role in the pathogenesis of PCO. High testosterone levels are associated with excess abdominal fat, obesity, and a higher risk of insulin resistance. Moreover, in both animal models and women with PCO, excess androgen leads to PCO morphology and ovulatory dysfunction [6, 7]. The Endocrine Society recommends using elevated total, bioavailable, or free serum testosterone levels in the diagnosis of PCO. However, O'Reilly et al. [13] reported that a high androstenedione level is associated with an excess of testosterone. They also suggested that measuring serum testosterone and androstenedione is a good predictor of metabolic risk in PCO. Patients with adrenal hyperandrogenism and elevated DHEAS have a better metabolic profile than patients with ovarian hyperandrogenism and elevated free testosterone. In PCO women, a high DHEAS/free testosterone or DHEAS/total testosterone ratio is associated with a favorable metabolic phenotype [14].

The distinction between different PCO phenotypes is crucial for disease severity prognosis, the outcome of fertility, and treatment planning. The role of ultrasound (US) in PCO diagnosis, including transabdominal and transvaginal US, is discussed in various studies, depending mainly on morphologic changes, ovarian volume, and the number of follicles [15,16,17,18,19,20,21,22,23,24]. Magnetic resonance imaging (MRI) is a valuable alternative for assessing follicular characteristics and ovarian morphology in PCO patients. MRI is a noninvasive imaging modality that has excellent anatomic resolution and is not affected by surrounding fat thickness in comparison with the US, especially in obese patients [25, 26]. Diffusion-weighted imaging (DWI) is a noninvasive technique that measures the water movement in the extracellular space. Hypercellularity, or an increased nuclear-to-cytoplasmic ratio, limits water diffusivity. The apparent diffusion coefficient (ADC) is a measure of biological tissue's specific diffusion capacity [27, 28]. Arterial spin labeling (ASL) is an MRI technique that vests the noninvasive measurement of perfusion at the tissue level. Ovarian enhancement and perfusion vary according to hormone status and the phase of the menstrual cycle. Currently, ASL is known to be helpful in cerebral blood flow evaluation, but the assessment of pelvic organs' blood flow by ASL is not yet known [29, 30]. To the best of our knowledge, there is no prior research in English literature to assess the use of ASL with DWI in PCO phenotyping.

The purpose of this work is to determine the effectiveness of ASL and DWI metrics in PCO diagnosis and differentiation between its different phenotypes, as well as correlating these metrics with laboratory measurements.

Methods

Subjects

Ethical approval of this prospective study was obtained from the local institutional review board. All patients provided their written informed consent. The study included patients who met the Rotterdam criteria [31] and underwent TVS, having disturbed menstruation, hirsutism, and/or infertility. Seventy-two patients fulfilled the inclusion criteria of our study, out of the 76 patients initially diagnosed with PCO based on the Rotterdam criteria [31]. The four excluded patients had a dominant follicle larger than 10 mm and bad imaging quality. The remaining 72 patients aged 23–43 years old. The patients were categorized into classic phenotypes, included type I (n = 36) and type II (n = 10), and non-classic form, which included type III (n = 16) and type IV (n = 10). Serum levels of testosterone and DHEAS were measured for all cases. The body mass index (BMI) was calculated for each case. Another 20-age-matched control group, aged 20–42 years, underwent ASL and DWI of the pelvis as well as measurement of serum levels of testosterone and BMI calculation.

MR imaging

The procedure of MR examination was explained to the patient. Patients fast for 4–6 h. Before MR imaging, an intravenous injection of 10 mg of an antispasmodic drug (Visceralgine) was given to reduce bowel peristalsis, particularly for diffusion-weighted imaging. MRI was done during the early follicular phase (about the 5th–6th days of the menstrual cycle). MR imaging was performed on a 1.5-T MR imaging unit (Ingenia, Philips, the Netherlands). All the patients were imaged in the supine position, head first, using a pelvic phased-array coil (Table 1).

Table 1 The sequences used in the study are listed in the following table
Full size table

ASL perfusion imaging was performed with a pseudo-continuous labeling technique. A single time point obtained after the labeling pulse with fast spin-echo single-shot echo-planar imaging parallels sequence. The positioning of the labeling pulse covered both ovarian arteries, and the post-labeling delay was 1.5 s. The parameters used were: repetition time/echo time 3600/5.9 ms, echo train length 21, matrix 128 × 96, FOV 240 × 240 mm2, imaging/tagging slice thickness 10/200 mm. The imaging of each inversion time was in 90 s.

Image analysis

The analysis of images was done blindly by two expert radiologists in female imaging over 15 and 10 years, respectively. In ASL imaging, the post-processing of arterial spin-tagging data includes initial subtraction of alternating tag and control image pairs, motion correction, and generating ASL gray-scale and colored maps. Regions of interest (ROIs) are manually positioned in the ovarian high-signal area within the ovarian stroma. Both regions' ROIs were comparable in size, with a range of 0.25 to 2.0 cm2. Images were compared with other conventional MR images to avoid follicles. In DWI, the application of matched ADC maps was done using functional tool software from Phillips Advantage windows Workstation. ADC values were measured by manual placement of ROIs of similar size and location within the stroma of both ovaries. Images were compared also with other conventional MR images to avoid follicles.

Statistical analysis

The IBM SPSS software package version 20.0 was used to analyze the data. Numbers and percentages were used to describe the qualitative data. Measures of quantitative data included mean, standard deviation, and range (minimum and maximum) after testing normality using the Shapiro–Wilk test. The significance of the obtained results was judged at the 5% level. All tests were two-tailed. The tests used were: Student t test (for quantitative parametric variables, to compare between different groups), the ROC curve (for detection of validity and cut-off point in comparison with sure diagnostic test), validity indices (sensitivity, specificity, accuracy), and lastly, binary logistic regression was used to detect the probability of combined variables in the prediction of cases with ROC curve application for saved probabilities.

Results

Seventy-two patients fulfilled the inclusion criteria of our study, out of the 76 patients initially diagnosed with PCO based on the Rotterdam criteria [31]. The four excluded patients had a dominant follicle larger than 10 mm and bad imaging quality. The remaining 72 patients aged 23–43 years old. The patients were classified into classic phenotypes, including type I (n = 36) and type II (n = 10) (Fig. 1, 2), and non-classic form, including type III (n = 16) and type IV (n = 10). The other control group included 20 females aged 20–42 years. Demographic details are provided (Table 2).

Fig. 1
figure 1

Classic Polycystic ovaries: A 26-year-old female with PCO A Axial T2-weighted image shows high signal small uniform peripherally arranged follicles, while the central stroma is of comparatively low T2. B, C ASL blood flow gray scale map B colored scale C show high signal intensity within ovarian stroma parts are 790.13, 812.02 mL/100 g/min for right and left sides, respectively. D and E DWI and ADC map show high signal of ovarian stoma on DWI image D with ADC values of ovarian stroma are 1.091, 1.012 × 10–3 mm2/s for right and left sides, respectively

Full size image
Fig. 2
figure 2

Non-classic Polycystic ovaries: A 22 years old female with PCO A Axial T2-weighted image shows high signal small uniform peripherally arranged follicles, while the central stroma is of comparatively low T2 B and C ASL blood flow gray scale map B colored scale C show high signal intensity within ovarian stroma are 720.45, 690.28 mL/100 g/min for right and left sides, respectively. D and E DWI and ADC map show high signal of ovarian stoma on DWI image D with ADC values of ovarian stroma are 1.301, 1.35 × 10–3 mm2/s for right and left sides, respectively

Full size image
Table 2 Demographic and laboratory data of patients and control groups
Full size table

The mean BF values of ovarian stroma in PCO patients were (722 ± 114 and 723 ± 114 mL/100 g/min) and in control were (510 ± 79 and 539 ± 11 mL/100 g/min) by first and second observers, respectively, with a significant difference (P = 0.001). Selection of (510.38 and 514.73 mL/100 g/min) as a threshold value of BF of ovarian stroma used for diagnosis of PCO revealed AUC of 0.94 and 0.89, accuracy of 96% and 92%, sensitivity of 98% and 95%, specificity of 90% and 80% by first and second observers, respectively. The mean ADC values of ovarian stoma in PCO patients were (1.20 ± 0.37 and 1.16 ± 0.34 × 10–3 mm2/s) and in control females were (1.66 ± 0.14 and 1.61 ± 0.23 × 10–3 mm2/s) for first and second observers, respectively, and a significant difference was observed (P = 0.001). Selection of (1.57 and 1.54 × 10–3 mm2/s) as a threshold of ADC of ovarian stroma for diagnosis of PCO yielded AUC of 0.85 and 0.84, accuracy of 79% and 80%, sensitivity of 80% and 81%, specificity of 75% and 75% for the first and second observers, respectively (Table 3). Combined use of BF and ADC of ovarian stroma in the diagnosis of PCO yielded AUC of (0.94 and 0.90), accuracy of (91% and 86%), sensitivity of (91% and 90%), and specificity of (90% and 75%) for the first and second observers, respectively (Table 4) and (Fig. 3).

Table 3 Mean and SD of BF, ADC of PCO and controls
Full size table
Table 4 ROC curve of BF and ADC of PCO patients and control
Full size table
Fig. 3
figure 3

ROC of BF, ADC, and combined parameters in patients versus controls: A BF = 0.95 and 0.89 in first and second observers, respectively B ADC = 0.85 and 0.84 in first and second observers, respectively. C Combined = 0.95 and 0.90 in first observer and second observers, respectively

Full size image

The mean BF values of the stroma in classic PCO patients were (790 ± 47 and 790 ± 46.88 mL/100 g/min) and of non-classic PCO patients were (601 ± 97 and 604 ± 100 mL/100 g/min) for first and second observers, respectively, with a significant difference (P = 0.001). Selection of (689 and 702.5 mL/100 g/min) as a threshold value of BF of ovarian stroma used for differentiation of classic from non-classic PCO yielded AUC of 0.921 and 0.908, accuracy of 91% and 91%, sensitivity of 98% and 96%, and specificity of 81% and 85% for the first and second observers, respectively. The mean ADC values of ovarian stoma in classic PCO patients were (1 ± 0.32 and 1 ± 0.27 × 10–3 mm2/s) and non-classic PCO patients (1.5 ± 0.3 and 1.4 ± 0.3 × 10–3 mm2/s) for the first and second observers, respectively, with a significant difference (P = 0.001). Selection of (1.4 and 1.3 × 10–3 mm2/s) as a threshold value of ADC of ovarian stroma for differentiation of a classic from non-classic PCO revealed AUC of 0.85 and 0.84, accuracy of 77% and 81%, sensitivity of 84% and 91%, and specificity of 69% and 65% for the first and second observers, respectively (Table 5). Combined use of BF and ADC of ovarian stroma in differentiating classic from non-classic PCO revealed AUC of (0.956 and 0.957), accuracy of (91% and 91%), sensitivity of (97% and 93%), and specificity of (80% and 88%) for first and second observers, respectively (Table 6) and (Fig. 4).

Table 5 Mean and SD of BF and ADC of PCO phenotypes
Full size table
Table 6 ROC curve of BF and ADC of PCO classic and non-classic phenotypes
Full size table
Fig. 4
figure 4

ROC of BF, ADC and combined parameters in classic versus non-classic: A BF = 0.92 in first observer and 0.91 in the second observer B ADC = 0.86 in first observer and 0.85 in second observer C combined = 0.95 in both observers

Full size image

There is no significant difference between classic and non-classic types when using DHEAS as a reference (P = 0.059). There was a significant difference between the classic, non-classic, and control groups when using testosterone as an indicator (P = 0.001). Also, BMI exhibits significant variability between the three groups (P < 0.001) (Table 2).

The relationship between MRI-derived parameters and laboratory findings showed a significant negative correlation between ADC and both serum testosterone and DHEAS levels (P = 0.001) for both observers. However, a non-significant negative correlation between ADC and BMI was noticed for the second observer. The relation between BF and both serum testosterone and DHEAS levels showed a significant positive correlation (P < 0.05). However, a non-significant positive correlation between BF and BMI was noticed for both observers.

Discussion

In the present study, we aimed to evaluate the feasibility and reproducibility of ASL technique in the assessment of ovarian blood flow in patients with polycystic ovaries. Studying vascular changes in the ovaries of PCO women may allow us to gain further insights into the underlying pathophysiology of the disease. ASL blood flow measurements can help in PCO identification and phenotypic differentiation.

Studies on ASL measurement have been focused on the brain thus far [29, 30]. But, to the best of our knowledge, no studies have been done on the application of ASL in PCO patient evaluation. Our study demonstrated that ovarian stroma in PCO patients exhibits significantly higher blood flow (P = 0.001) compared to control cases. A recent study used an intravoxel incoherent motion (IVIM) MRI model to assess both Brownian water diffusion and microcapillary perfusion separately in PCO patients and control groups. They found high values of pseudo-diffusion coefficient (D*) and perfusion fraction (f) in PCO patients, indicating increased stromal ovarian vascularity [32].

Other studies using Doppler ultrasound stated that in cases of PCO, there’s ovarian stromal hyperemia with lower arterial Pulsatility Index (PI) and Resistive Index values (RI). The reduced PI and RI values suggest the increased number and dilatation of ovarian stromal vessels in PCO [33, 34]. In accordance with our results, another study examined the mean signal intensity-time curve and found that PCO patients demonstrated greater and more rapid ovarian enhancement and wash-out. This enhancement behavior on the DCE-MR imaging examination may differ significantly from that of control subjects [35, 36].

As observed in this study, ADC values were lower in patients with PCO compared to control cases (P = 0.001). These results are comparable to a similar study that concluded that ADC measurements of the ovarian stroma are lower in patients with PCO than in control subjects (P = 0.001) and this might help improve the diagnosis of PCO [27, 28, 32]. Reduced diffusivity, in general, is related to increased cellularity, increased cell size, or microcirculation, which was discussed in many studies before [37,38,39]. According to a study by Papachroni et al. [40], the ovarian stroma of PCO cases may have lower ADC values due to the abnormal deposition of excess collagen in the polycystic ovarian stroma. Other studies attributed the decrease in ADC value to high levels of vascular endothelial growth factor (VEGF) that have been detected in the stroma of polycystic ovaries [40, 41]. Similar findings of reduced ADC values and increased levels of VEGF were found in hepatocellular carcinoma [42].

Phenotype differentiation in cases of PCO is essential for management, assessment of disease severity, and complication prognosis. This depends mainly on clinical, laboratory data, and ovarian morphology on ultrasound. In this study, we study the role of ASL and DWI metrics in the differentiation of classic and non-classic phenotypes of PCO. Our study results showed that ASL blood flow-derived metrics showed higher values in classic phenotypes (A and B) compared to non-classic types (C and D) (P = 0.001), while ADC values were lower in classic than non-classic cases (P = 0.001). This may be attributed to increased ovarian stroma and vascularity in classic than non-classic PCO cases. A study using ultrasound for phenotype differentiation revealed that the ovarian volume differed significantly between classic and non-classic phenotypes of PCO, with an accuracy of 75% [10].

The combined values of ASL blood flow and ADC of the ovarian stroma in the present study have shown a high significant difference between PCO cases and control cases. A study by Ozkok et al. [32] also found significantly reduced ADC values and increased IVIM MRI-derived perfusion parameters in PCO patients. The combined use of ASL blood flow and ADC of the ovarian stroma also found a highly significant difference between classic and non-classic PCO phenotypes. This will add great value to phenotype differentiation by MR imaging, which, as far as we know, has not been discussed before in English literature.

Our results revealed that there is a significant difference in testosterone level and BMI between the classic, non-classic, and control groups (P = 0.001). While the level of DHEAS was non-significant in differentiation between classic and non-classic phenotypes (P = 0.059). Our results showed agreement with Yuan et al. [14], who reported that there is a highly significant correlation between testosterone and BMI with different phenotype groups of PCO. On the other hand, there is partial agreement with O'Reilly et al. [13], who reported a positive correlation between high testosterone and PCO, and disagreement with our results regarding BMI.

Significant increase in levels of testosterone and DHEAS was associated with reduced ADC and increased BF in PCO patients. Other studies also found a significant correlation between Doppler indices and laboratory ovarian dysfunction markers [33, 43]. The correlation between imaging and laboratory biomarkers would prove the reliability of using these markers as predictors for PCO.

This study has a number of limitations. First, it included a relatively small number of patients. Further research in a larger cohort is recommended. Second, a 1.5-T scanner was used in this study. It is advised that further studies be done on 3-T machines for better assessment. Third, ASL results could be impacted by the degree of arterial stenosis; further studies are required using multi-parametric MR imaging combined with contrast MR angiography [36, 44] to evaluate arterial stenosis in future.

Conclusions

We came to the conclusion that PCO diagnosis and differentiation between its phenotypes can be aided by combining ASL and DWI-derived metrics of ovarian stroma. A significant correlation between those imaging metrics and serum levels of testosterone and DHEAS supports the use of ASL and DWI in the evaluation of PCO.