Search

Background: Placenta accreta spectrum (PAS) disorders are associated with high mortality and morbidity due to postpartum hemorrhage, hysterectomy, and organ injury, and a multidisciplinary team is required for an individualized case management. In this study, we assessed the diagnostic and prognostic accuracy of the most important ultrasonographic (US) and magnetic resonance imagining (MRI) markers for PAS disorders. Material and Methods: The study included 39 adult pregnant patients with at least one previous cesarean delivery and both US and MRI investigations for placenta previa evaluated at the tertiary maternity hospital ‘Cuza Voda’, Iasi, between 2019 and 2021. The following US signs were evaluated: intra-placental lacunae, loss of the retroplacental hypoechoic zone, myometrial thinning < 1 mm, bladder wall interruption, placental bulging, bridging vessels, and the hypervascularity of the uterovesical or retroplacental space. The MRI signs that were evaluated were intra-placental dark T2 bands, placental bulging, loss of the retroplacental hypointense line on T2 images, myometrial thinning, bladder wall interruption, focal exophytic placental mass, and abnormal vascularization of the placental bed. Results: The US and MRI signs analyzed in our study presented adequate sensitivities and specificities for PAS, but no sign proved to be a useful predictor by itself. The presence of three or more US markers for accretion was associated with a sensitivity of 84.6.6% and a specificity of 92.3% (p < 0.001). The presence of three or more MRI signs supplemented these results and were associated with a sensitivity of 92.3% and a specificity of 61.5% for predicting PAS (p < 0.001). Moreover, US and MRI findings were correlated with FIGO grading and severity of PAS. Conclusions: Even though no US or MRI finding alone can predict PAS with high sensitivity and specificity, our study proves that the presence of three or more imagistic signs could significantly increase the diagnostic accuracy of this condition. Furthermore, US and MRI could be useful tools for evaluating prognostic and perinatal planning.

(1) Background: Fetal growth restriction is a relatively common disorder in pregnant patients with thrombophilia. New artificial intelligence algorithms are a promising option for the prediction of adverse obstetrical outcomes. The aim of this study was to evaluate the predictive performance of a Feed-Forward Back Propagation Network (FFBPN) for the prediction of small for gestational age (SGA) newborns in a cohort of pregnant patients with thrombophilia. (2) Methods: This observational retrospective study included all pregnancies in women with thrombophilia who attended two tertiary maternity hospitals in Romania between January 2013 and December 2020. Bivariate associations of SGA and each predictor variable were evaluated. Clinical and paraclinical predictors were further included in a FFBPN, and its predictive performance was assessed. (3) Results: The model had an area under the curve (AUC) of 0.95, with a true positive rate of 86.7%, and a false discovery rate of 10.5%. The overall accuracy of our model was 90%. (4) Conclusion: This is the first study in the literature that evaluated the performance of a FFBPN for the prediction of pregnant patients with thrombophilia at a high risk of giving birth to SGA newborns, and its promising results could lead to a tailored prenatal management.

Background/Objectives: Prolonged prelabour rupture of membranes (PROMs), and the resulting inflammatory response, can contribute to the occurrence of adverse neonatal outcomes, especially for early-preterm neonates. This prospective study aimed to measure neonates’ inflammatory markers in the first 72 h of life based on ROM duration. The second aim was to examine the relationship between PROMs, serum inflammatory markers, and composite adverse neonatal outcomes after controlling for gestational age (GA). Methods: Data from 1026 patients were analyzed considering the following groups: group 1 (ROM < 18 h, n = 447 patients) and group 2 (ROM > 18 h, n = 579 patients). These groups were further segregated depending on the GA at the moment of membranes’ rupture into subgroup 1 (n = 168 patients) and subgroup 2 (at least 34 completed weeks of gestation, n = 858 patients). Multiple logistic regressions and interaction analyses adjusted for GA considering five composite adverse neonatal outcomes and predictors were employed. Results: PROMs and high c-reactive protein (CRP) values significantly increased the risk of composite outcome 1 occurrence by 14% (95%CI: 1.03–1.57, p < 0.001). PROMs and high CRP values increased the risk of composite outcome 5 by 14% (95%CI: 1.07–1.78, p < 0.001), PROM and leukocytosis by 11% (95%CI: 1.02–1.59, p = 0.001), and PROMs and high PCT values by 21% (95%CI: 1.04–2.10, p < 0.001). Conclusions: The combination of PROMs and high CRP values significantly increased the risk of all evaluated adverse composite outcomes in early-preterm neonates and should point to careful monitoring of these patients.

Background/Objectives: There is a constant need to improve the prediction of adverse neurodevelopmental outcomes in growth-restricted neonates who were born prematurely. The aim of this retrospective study was to evaluate the predictive performance of a three-layered neural network for the prediction of adverse neurodevelopmental outcomes determined at two years of age by the Bayley Scales of Infant and Toddler Development, 3rd edition (Bayley-III) scale in prematurely born infants by affected by intrauterine growth restriction (IUGR). Methods: This observational retrospective study included premature newborns with or without IUGR admitted to a tertiary neonatal intensive care unit from Romania, between January 2018 and December 2022. The patients underwent assessment with the Amiel-Tison scale at discharge, and with the Bailey-3 scale at 3, 6, 12, 18, and 24 months of corrected age. Clinical and paraclinical data were used to construct a three-layered artificial neural network, and its predictive performance was assessed. Results: Our results indicated that this type of neural network exhibited moderate predictive performance in predicting mild forms of cognitive, motor, and language delays. However, the accuracy of predicting moderate and severe neurodevelopmental outcomes varied between moderate and low. Conclusions: Artificial neural networks can be useful tools for the prediction of several neurodevelopmental outcomes, and their predictive performance can be improved by including a large number of clinical and paraclinical parameters.

(1) Background: Trial of labor after cesarean (TOLAC) can be associated with significant maternal and neonatal complications, and the aim of this retrospective study was to calculate the risks and probabilities of these complications in two tertiary maternity centers in Romania. (2) Methods: A total of 216 patients who attempted TOLAC were included in the study and were segregated into two groups, depending on TOLAC success. Medical records were assessed, and clinical data were used to determine the maternal and neonatal risks and complications, using multinomial logistic regression and postestimation predictions. (3) Results: Our data indicated that patients who had a failed TOLAC had significantly higher risks and probabilities of uterine rupture, either complete or incomplete; intensive care unit (ICU) admission; and emergency hysterectomy. The newborns of these mothers had significantly higher risks and probabilities of low Apgar score at birth, neonatal intensive care unit (NICU) admission, and invasive ventilation. (4) Conclusions: Failed TOLAC could lead to significant maternal and neonatal complications, and women who attempt this procedure should be monitored in a tertiary center where a multidisciplinary team and an emergency operating room are available.

(1) Background: Numerous variables could influence the risk of rectal cancer recurrence or metastasis, and machine learning (ML)-based algorithms can help us refine the risk stratification process of these patients and choose the best therapeutic approach. The aim of this study was to assess the predictive performance of 4 ML-based models for the prediction of local recurrence or distant metastasis in patients with locally advanced low rectal adenocarcinomas who underwent neoadjuvant chemoradiotherapy and surgical treatment; (2) Methods: Patients who were admitted at the first Oncologic Surgical Clinic from the Regional Institute of Oncology, Iasi, Romania were retrospectively included in this study between November 2019 and July 2023. Decision tree (DT), naïve Bayes (NB), support vector machine (SVM), and random forest (RF) were used to analyze imagistic, surgical, and pathological data retrieved from the medical files, and their predictive performance was assessed; (3) Results: The best predictive performance was achieved by RF when used to predict disease recurrence (accuracy: 90.85%) or distant metastasis (accuracy: 89.63%). RF was closely followed by SVM (accuracy for recurrence 87.8%; accuracy for metastasis: 87.2%) in terms of predictive performance. NB and DT achieved moderate predictive power for the evaluated outcomes; (4) Conclusions: Complex algorithms such as RF and SVM could be useful for improving the prediction of adverse oncological outcomes in patients with low rectal adenocarcinoma.

(1) Background: Prenatal care providers face a continuous challenge in screening for intrauterine growth restriction (IUGR) and preeclampsia (PE). In this study, we aimed to assess and compare the predictive accuracy of four machine learning algorithms in predicting the occurrence of PE, IUGR, and their associations in a group of singleton pregnancies; (2) Methods: This observational prospective study included 210 singleton pregnancies that underwent first trimester screenings at our institution. We computed the predictive performance of four machine learning-based methods, namely decision tree (DT), naïve Bayes (NB), support vector machine (SVM), and random forest (RF), by incorporating clinical and paraclinical data; (3) Results: The RF algorithm showed superior performance for the prediction of PE (accuracy: 96.3%), IUGR (accuracy: 95.9%), and its subtypes (early onset IUGR, accuracy: 96.2%, and late-onset IUGR, accuracy: 95.2%), as well as their association (accuracy: 95.1%). Both SVM and NB similarly predicted IUGR (accuracy: 95.3%), while SVM outperformed NB (accuracy: 95.8 vs. 94.7%) in predicting PE; (4) Conclusions: The integration of machine learning-based algorithms in the first-trimester screening of PE and IUGR could improve the overall detection rate of these disorders, but this hypothesis should be confirmed in larger cohorts of pregnant patients from various geographical areas.

(1) Background: SARS-CoV-2 infection during pregnancy could determine important maternal and fetal complications. We aimed to prospectively assess placental immunohistochemical changes, immunophenotyping alterations, and pregnancy outcomes in a cohort of patients with COVID-19; (2) Methods: 52 pregnant patients admitted to a tertiary maternity center between October 2020 and November 2021 were segregated into two equal groups, depending on the presence of SARS-CoV-2 infection. Blood samples, fragments of umbilical cord, amniotic membranes, and placental along with clinical data were collected. Descriptive statistics and a conditional logistic regression model were used for data analysis; (3) Results: Adverse pregnancy outcomes such as preterm labor and neonatal intensive care unit admission did not significantly differ between groups. The immunophenotyping analysis indicated that patients with moderate–severe forms of COVID-19 had a significantly reduced population of T lymphocytes, CD4+ T cells, CD8+ T cells (only numeric), CD4+/CD8+ index, B lymphocytes, and natural killer (NK) cells. Our immunohistochemistry analysis of tissue samples failed to demonstrate positivity for CD19, CD3, CD4, CD8, and CD56 markers; (4) Conclusions: Immunophenotyping analysis could be useful for risk stratification of pregnant patients, while further studies are needed to determine the extent of immunological decidual response in patients with various forms of COVID-19.

(1) Background: HELLP (hemolysis, elevated liver enzymes, and low platelets) syndrome is a rare and life-threatening complication of preeclampsia. The aim of this study was to evaluate and compare the predictive performances of four machine learning-based models for the prediction of HELLP syndrome, and its subtypes according to the Mississippi classification; (2) Methods: This retrospective case-control study evaluated pregnancies that occurred in women who attended a tertiary maternity hospital in Romania between January 2007 and December 2021. The patients’ clinical and paraclinical characteristics were included in four machine learning-based models: decision tree (DT), naïve Bayes (NB), k-nearest neighbors (KNN), and random forest (RF), and their predictive performance were assessed; (3) Results: Our results showed that HELLP syndrome was best predicted by RF (accuracy: 89.4%) and NB (accuracy: 86.9%) models, while DT (accuracy: 91%) and KNN (accuracy: 87.1%) models had the highest performance when used to predict class 1 HELLP syndrome. The predictive performance of these models was modest for class 2 and 3 of HELLP syndrome, with accuracies ranging from 65.2% and 83.8%; (4) Conclusions: The machine learning-based models could be useful tools for predicting HELLP syndrome, and its most severe form—class 1.