RESUMO
OBJECTIVES: The primary aim was the validation of benign descriptors (BDs), followed by Assessment of Different NEoplasia's of the adneXa (ADNEX) (when BDs cannot be applied), in a two-step strategy to classify adnexal masses in pregnancy. The secondary aim was to describe the natural history of adnexal masses in pregnancy. METHODS: Retrospective analysis of prospectively collected data of women with an adnexal mass on ultrasonography identified during pregnancy between 2017 and 2022. The study was conducted at Queen Charlotte's and Chelsea Hospital, UK. Relevant clinical and ultrasound data were extracted from the medical records and ultrasound software astraia. Adnexal masses were classified and managed according to expert subjective assessment (SA). Ultrasound features were recorded prospectively at the time of ultrasound examination. Borderline ovarian tumours (BOT) were classified as malignant. Benign Descriptors (BDs) were applied to classify adnexal masses, in cases where BDs were not applicable, the ADNEX model (using a risk of malignancy of >10%) was used, in a two-step strategy. The two-step strategy was applied retrospectively. The reference standard used was histology (where available) or expert SA at the postnatal ultrasound scan. RESULTS: 291 women with a median age of 33 (IQR 29-36) years presented with an adnexal mass in pregnancy, at a median gestation of 12 (IQR 8-17) weeks. 267 (267/291, 91.8%) women were followed up to the postnatal period, as 24 women (24/291, 8.2%) were lost to follow up. Based on the reference standard, 4.1% of adnexal masses (11/267) were classified as malignant (all BOTs) and 95.9% (256/267) as benign (41 on histology and 215 based on expert SA at postnatal ultrasound). BDs could be applied to 68.9% of adnexal masses (184/267); of these only one mass (BOT) was misclassified as benign (1/184, 0.5%). ADNEX was used to classify the residual masses (83/267) and misclassified three BOTs as benign (3/10, 30.0%) and 25 benign masses (based on reference standard) as malignant (25/73, 34.2%), 13 (13/25, 52.0%) of these were classified as decidualised endometriomas on expert SA, with confirmed resolution of decidualisation in the postnatal period. The two-step strategy had a specificity of 90.2%, sensitivity of 63.6%, negative predictive value of 98.3% and positive predictive value of 21.9%. 56 (56/267, 21.0%) women had surgical intervention, four as an emergency during pregnancy (4/267, 1.5%,) and four (4/267, 1.5%) electively during caesarean section. 48 (48/267, 18.0%) women had surgical intervention in the post-natal period, 11 (11/267, 4.1%) in the first 12 weeks postnatal and 37 >12 weeks (37/267, 13.9%) postnatal. 64 (64/267, 24.0%) adnexal masses resolved spontaneously during follow up. Cyst-related complications occurred in four women (4/267, 1.5%) during pregnancy (ovarian torsion n=2, cyst rupture n=2) and six (6/267, 2.2%) in the postnatal period (all ovarian torsion). 196 (196/267, 73.4%) had a persistent adnexal mass, including one of the women who had an ovarian torsion and underwent de-torsion and had a persistent adnexal mass at postnatal ultrasound. Presumed decidualisation occurred in 31.1% (19/61) of endometriomas and had resolved in 89.5% (17/19) by the first postnatal ultrasound scan. CONCLUSION: We found Benign Descriptors apply to most masses in pregnancy, however the small number of malignant tumours in the cohort (4.1%) restricted the evaluation of the ADNEX model, so expert subjective assessment should be used to classify adnexal masses in pregnancy, when BDs do not apply. A larger multicentre prospective study is required to evaluate the use of the ADNEX model to classify adnexal masses in pregnancy. Our data suggests that most adnexal masses can be managed expectantly during pregnancy given a large proportion of masses spontaneously resolved and the low risk of complications. This article is protected by copyright. All rights reserved.
RESUMO
AIM: To develop a multi-task learning (MTL) V-Net for pulmonary lobar segmentation on computed tomography (CT) and application to diseased lungs. MATERIALS AND METHODS: The described methodology utilises tracheobronchial tree information to enhance segmentation accuracy through the algorithm's spatial familiarity to define lobar extent more accurately. The method undertakes parallel segmentation of lobes and auxiliary tissues simultaneously by employing MTL in conjunction with V-Net-attention, a popular convolutional neural network in the imaging realm. Its performance was validated by an external dataset of patients with four distinct lung conditions: severe lung cancer, COVID-19 pneumonitis, collapsed lungs, and chronic obstructive pulmonary disease (COPD), even though the training data included none of these cases. RESULTS: The following Dice scores were achieved on a per-segment basis: normal lungs 0.97, COPD 0.94, lung cancer 0.94, COVID-19 pneumonitis 0.94, and collapsed lung 0.92, all at p<0.05. CONCLUSION: Despite severe abnormalities, the model provided good performance at segmenting lobes, demonstrating the benefit of tissue learning. The proposed model is poised for adoption in the clinical setting as a robust tool for radiologists and researchers to define the lobar distribution of lung diseases and aid in disease treatment planning.
