Artificial intelligence (AI) is revolutionizing how we practice medicine. In areas where we have traditionally struggled, such as diagnosing endometriosis, AI has significant potential to improve the breadth and accuracy of diagnostic services offering a great benefit to patient care. When developing AI models for diagnosis, the ‘ground truth’ refers to the reference standard used in the labelling of the data used to train the model. Conventionally, in clinical medicine, we correlate any new diagnostic tool to the established ‘gold standard’, which in the case of endometriosis is laparoscopic visualization of lesions and histological confirmation. This method however is increasingly recognized as imperfect. Acknowledgement of the limitations of surgery and recent improvements in the diagnostic capability of imaging technologies to detect endometriosis, has created a situation where endometriosis no longer has one clear ‘gold standard’ for diagnosis. In this commentary, we will explore the impact of this on AI-driven endometriosis diagnostic tools and propose novel ways this could be addressed in the context of creating ground truths for endometriosis diagnosis.

Endometriosis ultrasound reports are often unstructured free-text documents that require manual abstraction for downstream tasks such as analytics, machine learning model training, and clinical auditing. We present EndoExtract, an on-premise LLM-powered system that extracts structured data from these reports and surfaces interpretive fields for human review. Through contextual inquiry with research assistants, we identified key workflow pain points: asymmetric trust between numerical and interpretive fields, repetitive manual highlighting, fatigue from sustained comparison, and terminology inconsistency across radiologists. These findings informed an interface that surfaces only interpretive fields for mandatory review, automatically highlights source evidence within PDFs, and separates batch extraction from human-paced verification. A formative workshop revealed that EndoExtract supports a shift from field-by-field data entry to supervisory validation, though participants noted risks of over-skimming and challenges in managing missing data.

Background Rapid advances in transvaginal ultrasound techniques to detect endometriosis (eTVUS) require navigation of a steep learning curve. Artificial intelligence (AI) is playing an ever-increasing role in ultrasound and holds potential to help expedite eTVUS training. This study aimed to outline the range of eTVUS training currently being undertaken worldwide, identify barriers and enablers in performing eTVUS, and understand how AI-assisted tools could enhance self-learning of eTVUS. Method An online cross-sectional survey was disseminated to healthcare professionals who perform TVUS. A combination of multiple choice and free-text questions were asked regarding demographic information, experience performing eTVUS, training undertaken for eTVUS, experience with AI in ultrasound and how end-users believed AI could help with learning and performing eTVUS. Statistical and thematic analyses were performed. Results A total of 407 response, from 33 countries were included in the final analysis. Online self-directed learning (53.3%) was the most undertaken training method for eTVUS. Working with a skilled mentor was rated as the most helpful training method (mean 4.49 [range,3–5]), which was also the most desired, yet most difficult-to-access training method reported (42.3%). Inadequate training/education in the eTVUS technique (37.1%) and lack of confidence in recognising the appearance of endometriosis on ultrasound (37.1%) were the most common barriers. Training/education in eTVUS techniques was reported as the strongest facilitator to enable performing eTVUS (58.7%).Most respondents (64.7%) stated they had not used AI in their ultrasound workflow but 64.3% stated they would use an AI tool to assist with learning eTVUS, if available. Overwhelmingly, respondents wanted a teaching tool to help recognise disease (68.3%), which would ideally be built into an ultrasound machine (56.6%). Conclusion Self-directed online learning and reading was the most common form of training undertaken for eTVUS with difficulties reporting accessing skilled mentors as a significant barrier to eTVUS implementation. However, most health professionals would use AI tools for learning if available.

Women’s health research is essential for addressing gender-specific disparities and supporting equitable healthcare delivery. In South Australia, researchers have produced numerous evidence synthesis outputs, including systematic reviews, scoping reviews, and meta-analyses. However, there has been no comprehensive mapping of these studies to understand their thematic focus, methodological quality, and opportunities for improvement. This mapping review systematically examined women’s health evidence synthesis led by researchers affiliated with South Australian institutions. Following PRISMA-ScR guidelines and a registered protocol, a systematic search of ScienceDirect and grey literature sources identified 246 eligible studies. The most common themes were preconception, pregnancy, postpartum, and intrapartum health, while areas such as sexual health, mental health, Indigenous health, and violence and abuse were notably underrepresented. Methodological assessment revealed that over half of the studies were systematic reviews or meta-analyses, yet many lacked prospective registration, formal quality appraisal, or adherence to reporting standards. These findings highlight both the strengths and critical gaps in South Australia’s women’s health evidence synthesis landscape. Addressing these disparities through targeted funding, stronger methodological training, and adherence to review guidelines will be essential to improve research quality, guide policy development, and ensure that women’s health research effectively supports equitable health outcomes across the state.

In this study, we evaluate locally deployed large language models (LLMs) for converting unstructured endometriosis transvaginal ultrasound (eTVUS) reports into structured data. Across 49 de-identified reports, we compared three on-premise LLMs (7B/8B and 20B parameters) against expert human extraction using a 185-field schema. The 20B model achieved the highest mean accuracy (86.02%), substantially outperforming the smaller models. Crucially, LLMs and humans exhibited complementary error patterns: the LLM excelled on structured fields (date formatting, measurement decomposition) where humans made protocol errors, while humans demonstrated superior performance on interpretive fields involving negation and clinical terminology. Targeted prompt engineering yielded only marginal gains, indicating that these errors reflect model limitations rather than instruction gaps. These findings support a human-in-the-loop workflow in which the LLM generates structured drafts, automated validation flags rule-verifiable errors, and human review focuses on fields requiring clinical interpretation.

The efficacy of deep learning-based Computer-Aided Diagnosis (CAD) methods for skin diseases relies on analyzing multiple data modalities (i.e., clinical+dermoscopic images, and patient metadata) and addressing the challenges of multi-label classification. Current approaches tend to rely on limited multi-modal techniques and treat the multi-label problem as a multiple multi-class problem, overlooking issues related to imbalanced learning and multi-label correlation. This paper introduces the innovative Skin Lesion Classifier, utilizing a Multi-modal Multilabel TransFormer-based model (SkinM2Former). For multi-modal analysis, we introduce the Tri-Modal Cross-attention Transformer (TMCT) that fuses the three image and metadata modalities at various feature levels of a transformer encoder. For multi-label classification, we introduce a multi-head attention (MHA) module to learn multi-label correlations, complemented by an optimisation that handles multi-label and imbalanced learning problems. SkinM2Former achieves a mean average accuracy of 77.27% and a mean diagnostic accuracy of 77.85% on the public Derm7pt dataset, outperforming state-of-the-art (SOTA) methods.

Australia’s diverse migrant population necessitates a deeper understanding of migrant health needs, particularly in sexual and reproductive health. Female sexual dysfunction (FSD) significantly impacts quality of life, yet evidence on its prevalence and associated factors among migrant women in Australia is limited. This study aimed to explore FSD prevalence among migrant women from low and middle-income countries (LMICs) residing in Australia, compare FSD prevalence between migrant and Australian-born women, and examine socio-demographic factors associated with FSD in both groups. This national survey included reproductive-aged women (N = 868), comprising migrant women from LMICs (N = 421) and Australian-born women (N = 447). Participants were recruited through quota sampling via the Qualtrics online platform. The study employed the Female Sexual Function Index (FSFI) and a demographic questionnaire. Data analysis involved descriptive statistics, chi-square tests, and logistic regression. FSFI domain comparisons revealed significant differences between migrant and Australian-born women, migrant women reported significantly better overall sexual function (24.98 ± 7.18 vs. 23.57 ± 0.96, p = 0.01). Longer relationships were negatively associated with sexual function, while religious affiliation showed a significant impact on sexual dysfunction compared to no religious affiliation. Logistic regression analysis highlighted those high-income migrants had higher odds of better sexual function (OR: 2.27, 95% CI: 1.10-4.66) compared to low-income migrants. Regarding religion, non-religious Australian showed higher odds (OR: 3.24, 95% CI: 1.12-9.32) compared to religious ones. This pioneering study highlighted the need for tailored interventions considering socioeconomic status and cultural background, providing a foundation for further research on the intersections of migration, culture, and sexual well-being. This study contributes to a more comprehensive understanding of sexual health in Australia’s multicultural context, promoting overall well-being and quality of life for all women. No.

Introduction International migrants comprise 3.6% of the global population and face systemic barriers to accessing sexual and reproductive health (SRH) services, such as contraception, safe abortion care and sexual function support. In high-income countries, policy frameworks vary widely, with migration status significantly influencing entitlement and access to host countries. This protocol outlines a planned study to systematically analyse SRH policies in high-income countries with strong migrant integration frameworks, aiming to identify policy gaps, assess inclusivity and inform recommendations to strengthen Australia’s SRH policy landscape.Methods and analysis This study employs a systematic policy analysis using the Joanna Briggs Institute scoping review methodology. Countries with ≥10% migrant populations and a Migrant Integration Policy Index health score ≥70 will be included. 13 countries meet these criteria, including Australia, Canada and Sweden. A comprehensive search of academic databases (PubMed, Scopus, Web of Science, Cumulative Index to Nursing and Allied Health Literature and ProQuest Public Health) and grey literature from governmental and non-governmental sources will be conducted. Data extraction will follow Bacchi’s ‘What’s the Problem Represented to Be?’ approach. Thematic analysis will combine deductive and inductive methods to examine the extent to which SRH policies address migrant and refugee needs, including sexual function, safe abortion care and fertility care. A comparative policy matrix will identify strengths, limitations and best practices.Ethics and dissemination As this study analyses publicly available policy documents, ethics approval is not required. Findings will be disseminated through peer-reviewed publications and policy briefs targeting stakeholders involved in SRH policy and migrant health.Registration details This protocol is registered with the Open Science Framework (OSF): https://doi.org/10.17605/OSF.IO/AYZ6P

Objectives:Accurate diagnosis of pathology from ultrasound images is reliant upon images of a suitable diagnostic quality being acquired. This study aimed to create a novel machine learning model to automatically assess transvaginal ultrasound (TVUS) image quality for gynaecological ultrasound. Method:Six imaging professionals (two sonographers, two gynaecological sonologists and two radiologists) assigned a quality score to 150 TVUS images from 50 cases (50 uterus images and 100 ovary images). Images were given a score of 1-4 (1-reject/image inaccurate, 2-poor quality, 3-suboptimal quality or 4-optimal quality). As variation existed between the scores assigned by the labellers, we framed this problem as a multi-annotator noisy label problem. To address this, a new machine learning architecture was developed, combining a weighted ensemble algorithm to estimate consensus labels and a multi-axis vision transformer (MaxViT) to handle noisy labels, improving model accuracy in predicting image quality. Forty cases (120 images) were used for model training, while the remaining 10 cases (30 images) were reserved as a test set for model evaluation. Results:The novel machine learning architecture we created was able to successfully determine image quality with a validation accuracy of 80% and a macro average recall of 77%. This significantly improved upon the 57% accuracy of the baseline machine learning method (ResNet50). The MaxViTs were able to outperform human performance in most cases, with an accuracy of 80% surpassing four of the six human labellers. Conclusions:This novel machine learning model offers an automated method of assessing the quality of TVUS images. The tool has the potential to provide real-time feedback to those performing TVUS, reduce the need for repeated imaging, and improve the diagnosis of gynaecological pathology.

Abstract Study question Can we improve the diagnostic accuracy of the detection of POD obliteration in endometriosis magnetic resonance imaging, by leveraging results from unpaired eTVUS data sets? Summary answer We illustrate effective multimodal analysis methods improve POD obliteration detection accuracy from eMRI datasets, with an Area Under the Curve (AUC) from 65.0% to 90.6%. What is known already Traditionally, women investigated for pelvic pain and endometriosis, wait 6.4 years for laparoscopic diagnosis. There is a need for a more timely, non-invasive, accessible diagnostic tool. IMAGENDO is designed to combine eTVUS and eMRI using Artificial Intelligence (AI) to address this delay. We have previously demonstrated detection of pelvic endometriosis, including Pouch of Douglas (POD) obliteration, has a 95% specificity from endometriosis ultrasounds (eTVUS) and 72% from endometriosis magnetic resonance imaging (eMRI). This preliminary data has shown our novel multimodal AI approach, using imaging data from eTVUS and eMRIs, can improve diagnostic accuracy when detecting POD obliteration in endometriosis. Study design, size, duration The IMAGENDO study is a program of research designed to create a new diagnostic algorithm for endometriosis. The first part of the study describes the development of our initial algorithm using retrospective cross sectional transvaginal ultrasounds (n = 749), and magnetic resonance images (n = 89 private, n = 8984 public) from 9822 participants overall aged 18 to 45 years, collected between September 2011 and September 2022. Participants/materials, setting, methods Using public MRIs, we pre-trained a machine learning model, and fine-tuned the algorithm using private eMRIs to detect POD obliteration. Then unpaired eTVUSs were introduced to further improve our diagnostic model. We used a machine learning method known as Masked Autoencoder pretraining, which is unsupervised learning reconstructing masked data to generate a larger dataset. Then we embedded the data, compressing a large dataset into a small representation with the most salient features. Main results and the role of chance Scant training samples limited the generalisability of a 3D Vision Transformer to classify POD obliteration from MRI volumes, with an Area Under the Curve (AUC) of 65.0%. However, the masked auto-encoder pre-training partially mitigates this issue, improving the AUC to 87.2%. Adding knowledge distillation, and training a 3D Vision Transformer from scratch on such a small dataset is still challenging, with an AUC of 66.7%. However, adding both together: The knowledge distillation performance of 3D Vision Transformer with masked auto-encoder pre-training reaches an AUC of 77.2%, worse than without knowledge distillation, with an AUC of 87.2%. This could be due to the excessive domain shift between the pre-training dataset and TVUS dataset. However, fine-tuning the model from masked auto-encoder pre-training, the model improves accuracy from AUC=87.2% to AUC=90.6%. With all the steps using unmatched imaging from an alternative modality, this model ultimately demonstrated improvement in the AUC from 65% to 90% on our private MRI dataset. This is the first POD obliteration detection method that distils knowledge from TVUS to MRI using unpaired data, aiming to improve diagnostic accuracy of endometriosis from MRI; and the first machine learning method automatically detecting POD obliteration from MRI data to diagnose endometriosis. Limitations, reasons for caution Our eMRI datasets had some confounding problems, present as a result of artefacts, mislabelling, and misreporting. These were resolved using model checking, student auditing and expert radiology review. We will further test our algorithm with a diagnostic test accuracy study on at least two test cohorts. Wider implications of the findings Pre-training using digital data from different imaging modalities can improve the diagnosis of endometriosis especially when either imaging modality is missing. Provided specialist scanning is available, women with endometriosis will be able to obtain faster diagnosis prior to surgery. Trial registration number ACTRN12623000646640

Objective To identify gaps in existing evidence on preconception health interventions to improve the health outcomes of adolescents, young adults, and their offspring. Study design Evidence gap map (EGM) Methods Following the Campbell guidelines, we included reviews and interventional studies identified through searches on Medline and other electronic databases from 2010 to July 18th, 2023. Dual screening of titles/abstracts and full texts was conducted on Covidence software, followed by quality assessment and development of 2D-EGM using the EPPI-Reviewer and Mapper software. Results A total of 18 studies (124 papers) were identified, of which most of the studies were from higher- and upper-middle-income countries, with limited evidence from low-middle-income countries. More than half focused on females with limited evidence on men. The monitoring of adverse events of human papillomavirus (HPV) vaccination was the most well-evidenced area, with very little evidence on the herpes simplex virus candidate vaccine and other behavioural interventions. Perinatal outcomes were the most frequently reported outcomes followed by maternal and child health outcomes. Healthcare facilities (mostly clinical trials) were the most utilised delivery platforms, with limited or no evidence on communities, schools, and digital platforms. The overall quality of the systematic reviews was moderate while most of the trials had some concerns. Conclusion The study highlights a well-evidenced area in the safety of HPV vaccination with significant gaps in research on other key health interventions, particularly in non-healthcare settings. EGM suggests further research to evaluate the effectiveness of a broad range of preconception interventions, among adolescents and youth for improving long-term health outcomes.

BackgroundPreconception is the period before a young woman or woman conceives, which draws attention to understanding how her health condition and certain risk factors affect her and her baby’s health once she becomes pregnant. Adolescence and youth represent a life-course continuum between childhood and adulthood, in which the prepregnancy phase lacks sufficient research. ObjectiveThe aim of the study is to identify, map, and describe existing empirical evidence on preconception interventions that enhance health outcomes for adolescents, young adults, and their offspring. MethodsWe will conduct an evidence gap map (EGM) activity following the Campbell guidelines by populating searches identified from electronic databases such as MEDLINE, Embase, CINAHL, and Cochrane Library. We will include interventional studies and reviews of interventional studies that report the impact of preconception interventions for adolescents and young adults (aged 10 to 25 years) on adverse maternal, perinatal, and child health outcomes. All studies will undergo title or abstract and full-text screening on Covidence software (Veritas Health Innovation). All included studies will be coded using the Evidence for Policy and Practice Information (EPPI) Reviewer software (EPPI Centre, UCL Social Research Institute, University College London). Cochrane Risk of Bias tool 2.0 and Assessing the Methodological Quality of Systematic Reviews-2 (AMSTAR-2) tool will be used to assess the quality of the included trials and reviews. A 2D graphical EGM will be developed using the EPPI Mapper software (version 2.2.4; EPPI Centre, UCL Social Research Institute, University College London). ResultsThis EGM exercise began in July 2023. Through electronic search, 131,031 publications were identified after deduplication, and after the full-text screening, 18 studies (124 papers) were included in the review. We plan to submit the paper to a peer-reviewed journal once it is finalized, with an expected completion date in May 2024. ConclusionsThis study will facilitate the prioritization of future research and allocation of funding while also suggesting interventions that may improve maternal, perinatal, and child health outcomes. International Registered Report Identifier (IRRID)DERR1-10.2196/56052

Diagnosis of endometriosis has traditionally relied on laparoscopic surgery, which was considered the 'gold standard' diagnostic tool. This is not ideal as surgery carries risk, is expensive, is difficult to access, and disrupts patients work or education due to the recovery time needed. As such, imaging has been investigated as a potential method for non-invasive diagnosis, with transvaginal ultrasound showing high diagnostic accuracy for ovarian endometriomas and deep endometriosis. The advances in imaging capability led to recent international guidelines suggesting laparoscopy is no longer the 'gold-standard' for diagnosis and encouraging clinicians to utilise medical imaging as part of their diagnostic work-up for endometriosis. Imaging is emerging as not only a tool for planning endometriosis surgery but increasingly as the first approach for initial diagnosis. Given that transvaginal ultrasound is the primary imaging modality for assessment of gynaecological conditions, it is inevitable that sonographers will have a significant future role in endometriosis diagnosis. This moves away from endometriosis diagnosis being the exclusive realm of laparoscopic surgeons and increasingly involves medical imaging specialists. This review article will describe the origins of endometriosis ultrasound and the current capabilities of transvaginal ultrasound in this field. The expectations of sonographers in this evolving space will be explored, as well as recent novel research findings to gain insight into what the future of endometriosis diagnosis with ultrasound may look like.

BACKGROUND:Preconception health provides an opportunity to examine a woman's health status and address modifiable risk factors that can impact both a woman's and her child's health once pregnant. In this review, we aimed to investigate the preconception risk factors and interventions of early pregnancy and its impact on adverse maternal, perinatal and child health outcomes.METHODS:We conducted a scoping review following the PRISMA-ScR guidelines to include relevant literature identified from electronic databases. We included reviews that studied preconception risk factors and interventions among adolescents and young adults, and their impact on maternal, perinatal, and child health outcomes. All identified studies were screened for eligibility, followed by data extraction, and descriptive and thematic analysis.FINDINGS:We identified a total of 10 reviews. The findings suggest an increase in odds of maternal anaemia and maternal deaths among young mothers (up to 17 years) and low birth weight (LBW), preterm birth, stillbirths, and neonatal and perinatal mortality among babies born to mothers up to 17 years compared to those aged 19-25 years in high-income countries. It also suggested an increase in the odds of congenital anomalies among children born to mothers aged 20-24 years. Furthermore, cancer treatment during childhood or young adulthood was associated with an increased risk of preterm birth, LBW, and stillbirths. Interventions such as youth-friendly family planning services showed a significant decrease in abortion rates. Micronutrient supplementation contributed to reducing anaemia among adolescent mothers; however, human papillomavirus (HPV) and herpes simplex virus (HSV) vaccination had little to no impact on stillbirths, ectopic pregnancies, and congenital anomalies. However, one review reported an increased risk of miscarriages among young adults associated with these vaccinations.CONCLUSION:The scoping review identified a scarcity of evidence on preconception risk factors and interventions among adolescents and young adults. This underscores the crucial need for additional research on the subject.