Machine learning and radiomics for segmentation and classification of adnexal masses on ultrasound.

Ultrasound-based models exist to support the classification of adnexal masses but are subjective and rely upon ultrasound expertise. We aimed to develop an end-to-end machine learning (ML) model capable of automating the classification of adnexal masses. In this retrospective study, transvaginal ultrasound scan images with linked diagnoses (ultrasound subjective assessment or histology) were extracted and segmented from Imperial College Healthcare, UK (ICH development dataset; n = 577 masses; 1444 images) and Morgagni-Pierantoni Hospital, Italy (MPH external dataset; n = 184 masses; 476 images). A segmentation and classification model was developed using convolutional neural networks and traditional radiomics features. Dice surface coefficient (DICE) was used to measure segmentation performance and area under the ROC curve (AUC), F1-score and recall for classification performance. The ICH and MPH datasets had a median age of 45 (IQR 35-60) and 48 (IQR 38-57) years old and consisted of 23.1% and 31.5% malignant cases, respectively. The best segmentation model achieved a DICE score of 0.85 ± 0.01, 0.88 ± 0.01 and 0.85 ± 0.01 in the ICH training, ICH validation and MPH test sets. The best classification model achieved a recall of 1.00 and F1-score of 0.88 (AUC:0.93), 0.94 (AUC:0.89) and 0.83 (AUC:0.90) in the ICH training, ICH validation and MPH test sets, respectively. We have developed an end-to-end radiomics-based model capable of adnexal mass segmentation and classification, with a comparable predictive performance (AUC 0.90) to the published performance of expert subjective assessment (gold standard), and current risk models. Further prospective evaluation of the classification performance of this ML model against existing methods is required.

Clinical application of a nomogram based on age, serum FSH and AMH to select the FSH starting dose in IVF/ICSI cycles: a retrospective two-centres study.

OBJECTIVE: To externally validate a nomogram based on ovarian reserve markers as a tool to optimize the FSH starting dose in IVF/ICSI cycles. STUDY DESIGN: A two-centres retrospective study including 398 infertile women undergoing their first IVF/ICSI cycle (June 2013-June 2014). IVF data were retrieved from two independent IVF centres in Italy (San Raffaele Hospital, Centre 1; Verona Hospital, Centre 2). A central lab for the routine measurement of AMH and FSH was used for both centres. All women were treated based on physical and hormonal characteristics according to locally adopted protocols. The nomogram was then retrospectively applied to the patients comparing the calculated starting dose to the one actually given. RESULTS: In Centre 1, 64/131 women (48.8%) had an ovarian response below the target. While 45 of these patients were treated with a maximal FSH starting dose (≥225 IU), n=19/131 (14.5%) were treated with a submaximal dose. The vast majority of them (n=17/19) would have received a higher FSH starting dose by using the nomogram. Seventeen patients (n=17/131) had hyper response and about half of them would have been treated with a reduced FSH starting dose according to the nomogram. In Centre 2, 142/267 patients (53.2%) had an ovarian response below the target. While 136 of these were treated with a maximal FSH starting dose (≥225 IU), n=6/267 were treated with a submaximal dose. The majority of them (n=5/6) would have received a higher FSH starting dose. Thirty-two (n=32/267) patients had hyper response and more than half of them would have been treated with a reduced FSH dose. CONCLUSION: In both Centres, applying the nomogram would have resulted in more appropriate FSH starting doses compared to the the ones actually given based on clinicians choices. The use of an objective algorithm based on patient's age, serum FSH and AMH levels may thus be an effective advice on the selection of the tailored FSH starting dose. Hence, the use of this easily available nomogram could increase the proportion of patients achieving the optimal ovarian response.

Pregnancy outcome in women with endometriosis achieving pregnancy with IVF.

STUDY QUESTION: Are women with endometriosis who conceive with IVF at increased risk of preterm birth? SUMMARY ANSWER: Women with endometriosis who conceive with IVF do not face an increased risk of preterm birth. WHAT IS KNOWN ALREADY: The eutopic endometrium of women with endometriosis has been repeatedly shown to present molecular and cellular alterations. On this basis, it has been hypothesized that pregnancy outcome may be altered in affected women. However, to date, available evidence from epidemiological studies is scanty and conflicting. Data tended to be partly consistent only for an increased risk of preterm birth and placenta previa. STUDY DESIGN, SIZE, DURATION: Retrospective matched case-control study of women achieving an IVF singleton pregnancy progressing beyond 12 weeks' gestation. PARTICIPANTS/MATERIALS, SETTING, METHODS: Women achieving IVF singleton pregnancies that progressed beyond 12 weeks' gestation at two infertility units were reviewed. Cases were women with a history of surgery for endometriosis and/or with a sonographic diagnosis of the disease at the time of the IVF cycle. Controls were women without current or past evidence of endometriosis who were matched to cases by age (± 6 months), type of cycle (fresh or frozen cycle) and study period. Male factor and unexplained infertility were the most common diagnoses in the control group. Two hundred and thirty-nine women with endometriosis and 239 controls were selected. The main outcome of the study was the rate of preterm birth (birth < 37 weeks' gestation) regardless of the cause. Secondary analyses were performed for the most common obstetrical complications. MAIN RESULTS AND THE ROLE OF CHANCE: The rate of preterm birth was similar in the two study groups (14% and 14%, respectively, p = 0.89). The rate of live birth and the incidence of hypertensive disorders, gestational diabetes, small and large for gestational age newborns and neonatal problems also did not differ. In contrast, placenta previa was more common in women with endometriosis than controls (6% versus 1%, respectively; p = 0.006): The adjusted odds ratio was 4.8 (95% confidence interval: 1.4-17.2). LIMITATIONS, REASONS FOR CAUTION: As for all observational studies, confounders cannot be totally excluded. Moreover, the retrospective study design exposes the findings to some inaccuracies. For example, the independent role of adenomyosis could not be reliably assessed because this diagnosis is complex and would necessitate a prospective recruitment. Second, the selection of controls may also be a matter of concern because some affected women may have been erroneously included in this group. Third, even if the sample size is significant, it is insufficient for robust subgroup analyses. Finally, it is mandatory to point out that our conclusions are valid for IVF pregnancies only, and specific data from properly designed studies are required to support any inference for natural pregnancies. WIDER IMPLICATIONS OF THE FINDINGS: The results of our study suggest that women with endometriosis conceiving with IVF can be reassured regarding the risk of preterm birth. The observed association with placenta previa requires further investigation and may open a new avenue of research. STUDY FUNDING/COMPETING INTERESTS: No external funding was used for this study. None of the authors have any conflict of interest to declare.

Risk of miscarriage in women with endometriosis: insights from in vitro fertilization cycles.

OBJECTIVE: To evaluate whether women with endometriosis achieving singleton pregnancies with IVF face an increased risk of miscarriage. DESIGN: Matched case-control study. SETTING: Infertility units. PATIENT(S): Women achieving singleton pregnancies with the use of IVF were considered. Cases were women with a history of surgery for endometriosis and those who were documented the presence of ovarian endometriomas at the time of the IVF cycle (n = 313). Controls were matched to cases by age (±6 months), type of cycle (fresh or frozen cycle). and study period (n = 313). INTERVENTION(S): Retrospective review of women undergoing IVF. MAIN OUTCOME MEASURE(S): Rate of miscarriage before 12 weeks' gestation. RESULT(S): The number of miscarriages in women with and without endometriosis was similar, being 48 (15%) and 60 (19%), respectively. The odds ratio of miscarriage in affected women was 0.76 (95% confidence interval 0.50-1.16). The odds ratio adjusted for body mass index (BMI), parity, duration of infertility, and male factor infertility was 0.81 (95% confidence interval 0.53-1.25). Subgroup analyses according to the type of cycle, the number of embryos transferred, the presence of endometriomas, and the history of surgery for endometriosis did not document any subgroup at significant increased risk of miscarriage. CONCLUSION(S): The risk of miscarriage is not increased in women with endometriosis achieving pregnancy with the use of IVF.