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.

Beta-human chorionic gonadotrophin point of care testing for the management of pregnancy of unknown location.

RESEARCH QUESTION: Does a commercially available quantitative beta-human chorionic gonadotrophin (BHCG) point of care testing (POCT) device improve workflow management in early pregnancy by performing comparably to gold standard laboratory methods, and is the performance of a validated pregnancy of unknown location (PUL) triage strategy maintained using POCT BHCG results? DESIGN: Women classified with a PUL between 2018 and 2021 at three early pregnancy units were included. The linear relationship of untreated whole-blood POCT and serum laboratory BHCG values was defined using coefficients and regression. Paired serial BHCG values were then incorporated into the validated M6 multinomial logistic regression model to stratify the PUL as at high risk or at low risk of clinical complications. The sensitivity and negative predictive value were assessed. The timings required for equivocal POCT and laboratory care pathways were compared. RESULTS: A total of 462 PUL were included. The discrepancy between 571 laboratory and POCT BHCG values was -5.2% (-6.2 IU/l), with a correlation coefficient of 0.96. The 133 PUL with paired 0 and 48 h BHCG values were compared using the M6 model. The sensitivity for high-risk outcomes (96.2%) and negative predictive values (98.5%) was excellent for both. Sample receipt and laboratory processing took 135 min (421 timings), compared with 12 min (91 timings) when using POCT (P < 0.0001). CONCLUSIONS: POCT BHCG values correlated well with laboratory testing measurements. The M6 model retained its performance when using POCT BHCG values. Using the model with POCT may improve workflow and patient care without compromising on effective PUL triage.

Ambulatory human chorionic gonadotrophin (hCG) testing: a verification of two hCG point of care devices.

OBJECTIVES: Quantitative human chorionic gonadotropin (hCG) measurements are used to manage women classified with a pregnancy of unknown location (PUL). Two point of care testing (POCT) devices that quantify hCG are commercially available. We verified the i-STAT 1 (Abbott) and the AQT 90 FLEX (Radiometer) prior to use in PUL triage. METHODS: Tests for precision, external quality assurance (EQA), correlation, hook effect and recovery were undertaken alongside a POCT usability assessment during this prospective multi-center verification. RESULTS: Coefficients of variation ranged between 4.0 and 5.1 % for the three i-STAT 1 internal quality control (IQC) solutions and between 6.8 and 7.3 % for the two AQT IQC solutions. Symmetric differences in POCT EQA results when compared with laboratory and EQA stock values ranged between 3.2 and 24.5 % for the i-STAT 1 and between 3.3 and 36.9 % for the AQT. Correlation coefficients (i-STAT 1: 0.96, AQT: 0.99) and goodness of fit curves (i-STAT 1: 0.92, AQT: 0.99) were excellent when using suitable whole blood samples. An hCG hook effect was noted with the i-STAT 1 between 572,194 and 799,089 IU/L, lower than the hook effect noted with the AQT, which was between 799,089 and 1,619,309 IU/L. When hematocrit concentration was considered in sample types validated for use with each device, hCG recovery was 108 % with the i-STAT 1 and 98 % with the AQT. The i-STAT 1 scored lower on usability overall (90/130) than the AQT (121/130, p<0.001, Mann-Whitney). CONCLUSIONS: Both hCG POCT devices were verified for use in clinical practice. Practical factors must also be considered when choosing which device to use in each unit.

Clinical characteristics of functioning gonadotroph adenoma in women presenting with ovarian hyperstimulation: Audit of UK pituitary centres.

OBJECTIVE: Functioning gonadotroph adenomas (FGAs) are rare pituitary tumours stimulating ovarian function with potential life-threatening consequences in women. However, a lack of aggregated clinical experience of FGAs impairs management in affected women. The aim of this study is to present the clinical course of FGA-induced ovarian hyperstimulation syndrome (OHSS) cases as identified by some of the largest UK pituitary endocrine tertiary centres with a view to increasing awareness and improving diagnosis and management of women with FGA. DESIGN: A retrospective observational study; audit of eight UK regional pituitary centres for cases of FGAs. SETTING: Specialist neuroendocrine centres in the United Kingdom. PATIENTS AND MEASUREMENTS: Women diagnosed with FGA-induced OHSS. Description of their clinical course. RESULTS: Seven cases of FGA were identified in women, all causing OHSS. Mean age was 33.4 years at diagnosis. Abdominal pain, irregular periods, headache, and visual disturbances were reported at presentation by 100%, 71%, 57% and 43% of women, respectively. Three of seven women underwent ovarian surgery before FGA diagnosis. Six women underwent transsphenoidal surgery (TSS) with incomplete tumour resection in five of those, but all showed improvement or resolution in symptoms and biochemistry postoperatively. CONCLUSION: FGA is a rare cause of spontaneous OHSS. TSS improves clinical and biochemical features of ovarian hyperstimulation in FGAs. Improved awareness of FGA will prevent inappropriate emergency ovarian surgery.

Prognostic factors for post-traumatic stress, anxiety and depression in women after early pregnancy loss: a multi-centre prospective cohort study.

OBJECTIVES: To investigate prognostic factors for anxiety, depression and post-traumatic stress (PTS) symptoms 1 month after early pregnancy loss (EPL). DESIGN: A prospective cohort study. Consecutive women were recruited, and demographic and clinical data were collected. Surveys containing the Hospital Anxiety and Depression Scale (HADS) and Post-traumatic Stress Diagnostic Scale (PDS) were emailed 1 month after a loss. Univariable logistic regression was performed to link factors with caseness of anxiety, depression or PTS according to screening measures. SETTING: Early pregnancy units of three central London hospitals. PARTICIPANTS: 737/1116 eligible women with an EPL were recruited. 492 responded to HADS and 487 to PDS. PRIMARY AND SECONDARY OUTCOME MEASURES: Primary outcome is the area under the curve (AUC) to predict any psychological morbidity (defined as moderate/severe anxiety or depression, or meeting screening criteria for PTS) for each variable. Further outcomes are explained variation (R2) and p value for any morbidity, and AUC, explained variation, and p value for each morbidity separately. RESULTS: Women who had a past diagnosis of a psychiatric condition were more likely to meet criteria for anxiety, depression or PTS (75% for current diagnosis vs 55% for past vs 30% for no diagnosis; AUC 0.61; R2 8.4%; p<0.0001), as were those with previous pregnancy loss (48% vs 30%; AUC 0.59; R2 4.3%; p<0.0001). Most of the assessed factors did not demonstrate potential utility in predicting psychological distress, including gestational age, overnight admission, time taken for diagnosis, pre-existing children and the diagnosis itself (miscarriage vs ectopic vs other) (AUCs≤0.54; R2≤0.9%). CONCLUSIONS: Women with a history of mental health problems, or those with previous losses, may be at higher risk of psychological illness 1 month after pregnancy loss. However, prognostic ability was poor overall. All women should be considered at risk.

Evaluating the value of intrapartum fetal scalp blood sampling to predict adverse neonatal outcomes: A UK multicentre observational study.

OBJECTIVE: To evaluate the value of fetal scalp blood sampling (FBS) as an adjunct test to cardiotocography, to predict adverse neonatal outcomes. STUDY DESIGN: A multicentre service evaluation observational study in forty-four maternity units in the UK. We collected data retrospectively on pregnant women with singleton pregnancy who received FBS in labour using a standardised data collection tool. The primary outcome was prediction of neonatal acidaemia diagnosed as umbilical cord arterial pH < 7.05, the secondary outcomes were the prediction of Apgar scores<7 at 1st and 5th minutes and admission to the neonatal intensive care unit (NICU). We evaluated the correlation between the last FBS blood gas before birth and the umbilical cord blood and adjusted for time intervals. We constructed 2 × 2 tables to calculate the sensitivity, specificity, positive (PPV) and negative predictive value (NPV) and generated receiver operating curves to report on the Area Under the Curve (AUC). RESULTS: In total, 1422 samples were included in the analysis; pH values showed no correlation (r = 0.001, p = 0.9) in samples obtained within an hour (n = 314), or within half an hour from birth (n = 115) (r=-0.003, p = 0.9). A suboptimal FBS pH value (<7.25) had a poor sensitivity (22%) and PPV (4.9%) to predict neonatal acidaemia with high specificity (87.3%) and NPV (97.4%). Similar performance was noted to predict Apgar scores <7 at 1st (sensitivity 14.5%, specificity 87.5%, PPV 23.4%, NPV 79.6%) and 5th minute (sensitivity 20.3%, specificity 87.4%, PPV 7.6%, NPV 95.6%), and admission to NICU (sensitivity 20.3%, specificity 87.5%, PPV 13.3%, NPV 92.1%). The AUC for FBS pH to predict neonatal acidaemia was 0.59 (95%CI 0.59-0.68, p = 0.3) with similar performance to predict Apgar scores<7 at 1st minute (AUC 0.55, 95%CI 0.51-0.59, p = 0.004), 5th minute (AUC 0.55, 95%CI 0.48-0.62, p = 0.13), and admission to NICU (AUC 0.58, 95%CI 0.52-0.64, p = 0.002). Forty-one neonates had acidaemia (2.8%, 41/1422) at birth. There was no significant correlation in pH values between the FBS and the umbilical cord blood in this subgroup adjusted for sampling time intervals (r = 0.03, p = 0.83). CONCLUSIONS: As an adjunct tool to cardiotocography, FBS offered limited value to predict neonatal acidaemia, low Apgar Scores and admission to NICU.

p21-activated kinase 3 (PAK3) is an AP-1 regulated gene contributing to actin organisation and migration of transformed fibroblasts.

Activating Protein 1 (AP-1) plays a vital role in cell proliferation, differentiation and apoptosis. While de-regulation of AP-1 has been linked to many cancers, little is known regarding its downstream transcriptional targets that associate with cellular transformation. Previous studies identified PAK3, a serine/threonine kinase, as a potential AP-1 target gene. PAK3 has been implicated in a variety of pathological disorders and over-expression of other PAK-family members has been linked to cancer. In this study, we investigate AP-1 regulation of PAK3 expression and the role of PAK3 in cJun/AP-1-associated cellular transformation. Our results showed elevated PAK3 expression at both the mRNA and protein level in cJun-over-expressing Rat1a fibroblasts, as well as in transformed human fibroblasts. Elevated PAK3 expression in cJun/AP-1 over-expressing cells associated with a significant increase in PAK3 promoter activation. This increased promoter activity was lost when a single putative Jun binding site, which can bind AP-1 directly both in vitro and in vivo, was mutated. Further, inhibition of PAK3 using siRNA showed a regression in the cell morphology, migratory potential and actin organisation associated with AP-1 transformed cells. Our study is a first to describe a role for AP-1 in regulating PAK3 expression and suggest that PAK3 is an AP-1 target required for actin organization and migration observed in transformed cells.