RPS 1007 - Imaging in pregnancy and female tumours

Purpose:

To evaluate the ADNEX MR scoring system accuracy in predicting ovarian tumour malignancy.

Methods and materials:

A retrospective study was conducted on the basis of the analysis of pelvic MRI results of patients with ovarian lesions sent for clarifying diagnosis after ultrasound from 2018 to 2019, after excluding incomplete study protocols and observations without diagnosis verification. All ovarian lesions were analysed using the ADNEX MR scoring system algorithm.

Results:

The study included 118 patients (average age 48.6 (26-81)) and 136 lesions: 33 (24.3%) were malignant and 7 (21.2%) were borderline epithelial sex cord stromal tumours with a low degree of malignancy. The distribution of malignant tumours among the studied groups showed that lesions with a score of 2 were found in 1.37% (1/73), with a score of 3 were found in 11.5% (3/26), with a score of 4 were found in 62.5% (10/16), and with a score of 5 were found in 90.5% (19/21). The ADNEX MR scoring system demonstrated high accuracy: a sensitivity of 91.3% (95% confidence interval [CI], 86.1% -94.8%) and specificity of 96.7% (95% CI, 92.3% -98.6%) in the malignancy prediction.

Conclusion:

ADNEX MR scoring system is an informative method of ovarian tumour malignancy prediction, therefore implementation in the algorithm of preoperative diagnosis can be important for optimising the management tactics. The greatest difficulties in assessing arose when characterising borderline epithelial mucinous tumours and non-epithelial tumours with a low degree of malignancy.

Limitations:

A retrospective, single-institution study.

Ethics committee approval

Ethics committee approval was obtained.

Funding:

The study was funded by the Grant of the President of the Russian Federation.

Purpose:

To evaluate the predictive value and accuracy of individual MRI findings in the diagnosis of abnormally invasive placenta (AIP).

Methods and materials:

A retrospective PACS-based search was done for all the MRIs performed over 5 years in a tertiary referral centre for evaluation of possible AIP. The normal outcome was defined as an eventless vaginal delivery or cesarian section without the need for transfusion or hysterectomy. AIP was considered positive if pathologic prove was present. Patients with complicated deliveries but no hysterectomy or pathologic proof were excluded from the study. Individual findings evaluated included heterogeneity, dark placental bands, rounded edges, lumpy margins, and bulging uterine outline. Chi-square was used to compare findings. Sensitivity, specificity, PPV, and NPV for each individual finding were determined.

Results:

Over a period of 2 years, 392 MRIs were done to further evaluate for possible AIP, of which 152 cases had a final diagnosis of AIP and 38 were lost to follow-up. Sensitivity, specificity, and overall accuracy of MRI for diagnosis of AIP were 93, 90%, and 89%. Dark bands were most accurate (sensitivity, specificity, and overall accuracy of 91, 78%, and 82%). Lumpy margins and loss of retroplacental hypo intense band were least accurate. A combination of retroplacental dark bands, anteroinferior low-lying/previa, and bulging uterine margins were the most accurate combination, slightly increasing the sensitivity, specificity, and overall accuracy to 93, 92, and 91% (based on a logistic regression method).

Conclusion:

Not all the individual signs of AIP have the same diagnostic value in predicting the diagnosis. The single most useful determinant for the diagnosis was dark placental bands. The most useful combination of findings were a history of previous C/S, placenta previa, and dark placental bands. Adding other findings did not significantly raise the accuracy.

Limitations:

n/a

Ethics committee approval

n/a

Funding:

No funding was received for this work.

Purpose:

To evaluate differences in placental volume and vasculature between normal and FGR pregnancies by three-dimensional ultrasound (3D-USG).

Methods and materials:

703 pregnant women with singleton foetus at gestation age (GA) of 6-9 weeks (GA confirmed through transvaginal ultrasound) were enrolled. At 28-36 weeks, 3D-USG assessment of placental volume (PV), vascularisation index (VI), flow index (FI), vascularisation flow index (VFI), and an estimated foetal weight (EFW) was done. Those with EFW < 10^th percentile for GA were placed in the FGR group and the rest in the normal group. Follow-up till delivery was done in 121 FGR and 153 normal pregnancies. The diagnosis was confirmed through birth-weight in 107 and 111 women, respectively. Receiver operator characteristic (ROC) curves were plotted; correlation coefficients of PV, VI, FI, VFI with EFW were calculated. A Student t-test and Mann-Whitney U test were used.

Results:

A mean of PV, VI, and VFI were significantly lower (p<0.001) in FGR than normal (PV: 192.67 ± 24.07 vs 275.08 ± 94.82, VI: 3.26 ± 1.48 vs 10.53 ± 7.94, VFI: 2.44 ± 1.61 vs 5.46 ± 3.07) while FI had no significant difference between the two (FI: 46.41 ± 7.70 vs 49.19 ± 7.08, p=0.07). Area under ROC curve (AUC) were VI: 0.860, VFI: 0.799, PV: 0.791, and FI: 0.690. Sensitivity and specificity of cut-off values for PV (208 ml), VI (5.0), and VFI (3.2) were 74% and 69%, 91% and 75%, and 75% and 71%, respectively. PV showed a maximum correlation with EFW; 0.829 for FGR and 0.911 for normal pregnancies (p<0.05 each).

Conclusion:

3D-USG assessment of placental volume and vasculature serves as a complementary tool to EFW in discriminating FGR from normal pregnancies.

Limitations:

A tertiary hospital-based study; results cannot be generalised at a community level.

Ethics committee approval

n/a

Funding:

No funding was received for this work.

Purpose:

Sonosalpingography (SSG) has been a less commonly used tool for assessing the patency of fallopian tubes in subfertile females. Its significance is undermined by laparoscopic evaluation (LE) that allows simultaneous therapeutic procedures to restore its patency if the obstruction exists. But LE is invasive and expensive.

Methods and materials:

We performed this study with the aim of evaluating the role of SSG in the diagnosis of tubal obstruction, its cause, and relieving the obstruction.

50 subfertile females with a normal-appearing uterus and ovaries on transvaginal ultrasonography were included in our study. SSG was performed with an infusion of normal saline to evaluate tubal patency by recording free peritoneal spill. If peritoneal spill was absent bilaterally then the patient underwent laparoscopic evaluation. However, if unilateral or bilateral peritoneal spill was noted, then the patient was recruited for assisted reproductive techniques (ART) and the results were correlated with pregnancy.

Results:

Out of 50 patients, SSG demonstrated free peritoneal spill unilaterally in 46 patients who conceived with ART during the following six months. In the remaining 4 patients with a lack of bilateral spill on SSG, 2 revealed unilateral partial block and 2 revealed bilateral tubal block on LE. In all patients, SSG correctly depicted the site of obstruction. In 9 patients, it revealed PID (tubercular) by demonstrating flimsy peritubal adhesions and in 14 patients, higher pressure exerted during SSG restored the patency with sharp abdominal pain.

Conclusion:

Our study reveals that SSG is not only diagnostic but may also be a therapeutic tool. SSG should be used more often and can be used to at least segregate patients who prudently need laparoscopic evaluation.

Limitations:

n/a

Ethics committee approval

n/a

Funding:

No funding was received for this work.

Purpose:

To assess the accuracy of non-contrast pelvic magnetic resonance imaging (MRI) for the characterisation of adnexal masses and evaluate the reproducibility of a new non-contrast adnex MRI scoring system used in a tertiary referral centre.

Methods and materials:

Pelvic MRI examinations performed for adnexal mass characterisation or follow-up of endometriomas between the years 2008-2018 in our institution were retrospectively reviewed. In total, 484 patients with an adnexal mass who were either treated surgically or followed-up for at least one year were identified. After exclusion of patients without diffusion-weighted imaging (DWI), the final cohort included 297 patients with 356 adnexal masses. A randomised set of 100 patients was chosen from this cohort to assess the reproducibility of the proposed five-category scoring system. Various MRI features were assessed in both sets.

Results:

There were 303 benign and 53 malignant tumours in the study cohort. ROC analysis of the used scoring system for predicting overall outcome showed an AUC of 0.93. A cut-off score of ≥4 was associated with malignancy with an accuracy of 93.82% and PLR of 18 for predicting malignancy, with a sensitivity of 84.91% and a specificity of 95.38%. In the analysis of MRI features of the reproducibility set, inter and intraobserver agreement was moderate-almost perfect for most of the features and the repeatability of the scoring system.

Conclusion:

Non-contrast MRI scoring for adnexal mass characterisation seems to be promising with high accuracy and reasonable inter and intraobserver agreement. The validity of the scoring needs to be tested in multicentre prospective studies.

Limitations:

A single-centre, retrospective study.

Ethics committee approval

Service evaluation approval is present.

Funding:

No funding was received for this work.

Purpose:

To investigate the association between CT imaging traits and texture metrics with proteomic data in patients with high-grade serous ovarian cancer (HGSOC).

Methods and materials:

This retrospective, hypothesis-generating study included 20 patients with HGSOC prior to primary surgery. Two readers independently assessed the contrast-enhanced computed tomography (CT) images and extracted 33 imaging traits. All sites of suspected HGSOC were manually segmented and grey-level correlation and matrix-based texture features were computed. 3 texture features that represented intra-and inter-site tumour heterogeneity were used for analysis. An integrated analysis of transcriptomic and proteomic data identified proteins with a conserved expression between primary tumour sites and metastasis. Correlations between protein-abundance and various CT features were assessed using the Kendall tau rank correlation coefficient and the Mann-Whitney U test, whereas the area under the receiver-operating characteristic-curve (AUC) was reported as a metric of the strength and the direction of the association.

Results:

8 proteins were identified as significantly associated with CT-based imaging traits, with the strongest correlation observed between the MRI1 protein and peritoneal disease in the liver/right upper quadrant (P<0.001, AUC=0.940). The abundance of 4 proteins was associated with texture features that represented intra and inter-site tumour heterogeneity, with the strongest correlation between the GSTM1 protein and cluster dissimilarity (p=0.024, =0.368).

Conclusion:

This study provides the first insights into the potential associations between standard-of-care CT imaging traits and texture measures of intra-and intersite heterogeneity and the abundance of several proteins.

Limitations:

A small sample, retrospective study.

Ethics committee approval

This was a multi-institutional, institutional review board-approved, and Health-Insurance Portability and Accountability Act (HIPAA)-compliant study.

Funding:

The Mark Foundation for Cancer Research and the Cancer Research UK Cambridge Centre (C9685/A25177).

U.S. Department of Defense - Uniformed Services University of the Health Sciences (HU0001-16-2-0006).

National Cancer Institute (HHSN261200800001E).

Purpose:

To investigate the potential of the radiomic approach to MRI to the predict pathological complete response to treatment (pCR) in patients with locally advanced cervical cancer (LACC). The second goal was to develop and validate this MRI signature, to apply the model out of the habitat where it was created.

Methods and materials:

183 patients with LACC (FIGO IB2-IVA) treated in two hospitals were enrolled, 156 patients for the training set and 27 for the validation set, respectively. All patients underwent neoadjuvant chemoradiation therapy followed by radical surgery.

For each patient, a staging 1.5T MR image was acquired. 230 radiomic features belonging to 4 families (fractal, statistical, textural, and morphological features) were extracted. Considering the application of all the filters used, 1,889 features were extracted. Features selection and model training were carried out following an iterative method, ad-hoc developed for this study. Model selection was carried out using the area under the curve (AUC) of the receiving operator characteristic (ROC) curve as a target metric.

Results:

15 features were selected. The model showing the highest performance was a random forest with default parameters (RF_DEF). This model shows an AUC of 0.76 on the training and 0.82 on the external validation set.

Conclusion:

Our radiomic model showed a good predictive value with a high AUCs both in the training and validation group, so it is able to predict the probability of pathological complete response in patients undergoing chemoradiation therapy using pre-treatment images.

The presented model is able to predict patients’ response to treatment before commencement and use this information to support the clinical decision-making process.

Limitations:

Further external validation cohort should be studied.

Ethics committee approval

Institutional ethics committee approved the study.

Funding:

No funding was received for this work.

Purpose:

To evaluate the influence of contraceptives on the variability of quantitative kurtosis parameters of the ovaries over the menstrual cycle in diffusion-weighted magnetic resonance imaging (DWI).

Methods and materials:

This IRB-approved prospective study included 30 volunteers with or without oral-contraceptives (n=15 for each group, mean 27.9y) from 07/2017-09/2019. MRI was performed using a 1.5T MRI (Aera) with a multi-b-value-epi-DWI 0,50, 100, 800, 1500, and 2000s/mm² at three timepoints during the menstrual cycle (T1=day 1-5, T2=day 7-12, and T3=day 19-24). Ovaries were 3D-segmented (b=800s/mm²). Diffusion-kurtosis (K_app) was calculated. Quantitative parameters and coefficient-of-variation (CoeV) were analysed and compared between the groups using Students t-test after Shapiro-Wilk and Brown-Forsythe pretest.

Results:

Amongst the women without oral contraceptives, the CoeV for kurtosis parameters was significantly elevated in comparison to the group with oral contraceptives at Timepoint1 (0.34 vs 0.30;p=0.04) and Timepoint2 (0.36 vs 0.32,p=0.018) of the menstrual cycle, however not at Timepoint3 (0.29 vs 0.30;p=0.28). Mean absolute K_app was not significantly different in between the women with and without oral contraceptives for Timepoint1=0.618±0.057 vs 0.637±0.039], p=0.14, and Timepoint2=0.606±0.062 vs 0.616±0.059, p=0.54, however, at Timepoint3, significant differences were observed (0.649±0.058 vs 0.596±0.062, p=0.0013).

Conclusion:

Quantitative diffusion kurtosis imaging demonstrated a significantly higher coefficient of variation in women with a natural menstrual cycle compared to women using oral contraceptives in the follicular phase. The absolute mean values between the groups only differed significantly during the luteal phase. These findings should be considered when quantitatively applying kurtosis DWI in imaging of the ovaries.

Limitations:

A limited number of volunteers, no intra-individual repeated measures, and the influence on suspicious lesions was not investigated.

Ethics committee approval

IRB approved.

Funding:

No funding was received for this work.

Purpose:

To build a nomogram incorporating both clinical and radiomic features of the placenta on T2WI to predict the risk for PPH in pregnancies during caesarean delivery (CD).

Methods and materials:

207 pregnant women from hospital #1 and 91 pregnant women from hospital #2 were enrolled as training and validation cohorts, respectively. These women were suspected of having placenta accreta spectrum (PAS) disorders and underwent MRI for placenta evaluation. PPH was defined as more than 1,000 mL estimated blood loss (EBL) during CD. Radiomic features were selected based on their correlations with EBL. Radiomic, clinical, radiological, clinicoradiological, and clinicoradiomic models were built to predict the risk of PPH for each patient. The model with the best prediction performance was validated with its clinical application.

Results:

35 radiomic features showed a strong correlation with EBL. The clinicoradiomic model resulted in the best discrimination ability for risk prediction of PPH, with AUC of 0.888 (95% CI, 0.844-0.933) and 0.832 (95% CI, 0.746-0.913), sensitivity of 91.2% (95% CI, 85.8%-96.7%) and 97.6% (95% CI, 92.7%-100%) in the training and validation cohort, respectively. For patients with severe PPH (EBL more than 2,000 mL), 53 out of 55 pregnancies (96.4%) in the training cohort and 18 out of 18 (100%) pregnancies in the validation cohort were identified by the clinicoradiomic model.

Conclusion:

The clinicoradiomic model incorporating both prenatal clinical factors and radiomic signature of the placenta on T2WI showed good performance for risk prediction of PPH. The predictive model can identify severe PPH with high sensitivity.

Limitations:

The pathological diagnosis of PAS disorders was lacking in our study.MRI is not as widely used for clinical use in pregnant women as US.

Ethics committee approval

n/a

Funding:

This study has received funding from National Key R&D Program (2017YFE0103600).