Introduction: Endometriosis is a common benign gynaecological condition. However, malignant transformation to endometriosis-associated ovarian cancer (EAOC), though rare, presents a significant diagnostic challenge. Patients typically present with known ovarian endometriomas and may demonstrate new or evolving clinical symptoms, including worsening pelvic pain, increasing mass size, or atypical imaging appearances. Accurate sonographic differentiation between benign endometriosis and malignant transformation remains clinically challenging yet essential for timely management.
Method: This case series reviews pelvic ultrasound examinations of patients with known or suspected endometriosis who demonstrated atypical or concerning imaging features. Ultrasound assessment incorporated pattern recognition, grey-scale morphology, Doppler evaluation of vascularity, and comparison with prior imaging where available. Findings were interpreted in reference to established IOTA and O-RADS criteria.
Results: Key sonographic findings across cases included the development of solid components within endometriomas, papillary projections measuring ≥3 mm, internal vascularity, irregular cyst walls, and rapid interval growth. These imaging features prompted further investigation or surgical referral. The observed sonographic appearances were consistent with features described in the literature as suspicious for, or subsequently confirmed to be, malignant transformation.
Conclusion: This case series highlights the critical role of ultrasound in identifying features suggestive of EAOC when a structured and comparative imaging approach is applied. The findings support existing evidence that papillary projections, vascularised solid components, and architectural change are key sonographic indicators of malignancy.
Take home message: In patients with endometriosis, the presence of new solid, vascularized components, or morphological changes should prompt consideration of malignant transformation and appropriate escalation of care.

Transvaginal ultrasound (TVUS) is a first-line diagnostic method for deep endometriosis (DE). It, however, requires expertise. Typical ultrasound features of DE include hypoechoic, scarce, solid nodules affecting uterosacral ligaments, rectovaginal septum, posterior vaginal fornix, urinary bladder wall, or bowel often accompanied by adhesions between the organs in the pelvic, leading to obliteration
TVUS has demonstrated high pooled sensitivities (around 91%) and specificities (around 98%) for detection of DE in the bowel and high specificity (often > 90%) and good sensitivity (around 75-85%) for DE in other pelvic sites.
The lecture will cover ultrasound diagnostics of deep endometriosis in the pelvic including ultrasound appearance of DE in different locations, systematic scanning based in IDEA consensus, tips and tricks how to search for DE lesions in particular locations.

Introduction: Extraction of data from ultrasound reports is an essential step in clinical ultrasound research. However, it is a labour-intensive, time-consuming task. Large language models (LLMs) are artificial intelligence tools which hold capability to process large amounts of language data in seconds. In this study, we aimed to evaluate three locally deployed LLMs to convert unstructured endometriosis transvaginal ultrasound (eTVUS) reports into structured data for research synthesis.
Methods: Forty-nine eTVUS reports were extracted using three LLMs (Llama3-8b, Mistral-7b and GPT-oss:20b) into a structured excel spreadsheet to document the variables. The LLMs were hosted on a local server to ensure data security. The same reports were also extracted by a human research assistant utilising the same data sheet. An expert sonographer then reviewed all data extraction to assess for errors.
Results: Our human extractor achieved an accuracy of 98.40% (SD 2.13%). This exceeded the performance of all LLMs which showed accuracies of 86.02% (SD 6.87%)[GPT-oss: 20b), 80.53% (SD 4.58%) [Llama3-8b] and 78.89%(4.68%) [Mistral-7b]. The LLMs outperformed the human extractor in numeric fields whereas the human extractor outperformed the LLMs on text-based and categorical fields.
Conclusion/Take home message: Although LLM accuracy did not exceed that of human extraction, the model produced complementary error patterns that support a human-in-the-loop workflow. By automating routine structuring tasks and flagging potential inconsistencies, LLMs have the potential to substantially reduce manual reporting time, allowing imaging specialists to focus on higher-level semantic validation and clinical decision-making.

Background: Transvaginal ultrasound (TVUS) is an intimate diagnostic procedure requiring informed, voluntary consent. Professional guidelines mandate clear, patient-centred consent processes; however, anecdotal evidence suggests that consent for TVUS is sometimes rushed, inconsistent, or insufficiently trauma-informed and culturally sensitive. Limited published literature exists on patient experiences of the consent process for TVUS within the sonography profession. Gaps in consent practice may contribute to patient distress, loss of trust, avoidance of care, and ethical or legal risk.
Methods: This clinical audit, funded by the ASA, uses an anonymous online survey in a descriptive cross-sectional design. Adult outpatients undergoing TVUS at SAMI, Repat Health Precinct, SA Health, will be invited to participate between February and August 2026, with follow-up analysis and re-audit planned. The survey aims to evaluate patient-reported experiences of the consent process, focusing on clarity of information, communication of risks, opportunity to decline or withdraw consent, availability of a support person, cultural sensitivity, trauma-informed communication, and overall patient-centredness.
Results: Initial early results of the audit will be presented to help identify strengths and gaps in alignment with ethical standards, clinical guidelines, and best practice principles of trauma-informed and culturally competent care.
Conclusion: Evaluating patient experiences of TVUS consent provides critical insight into whether current practices uphold autonomy, emotional safety, and cultural humility. Findings will inform quality improvement initiatives and support consistent, ethical consent processes.
Take-home message: Strengthening trauma-informed, culturally sensitive consent processes can reduce patient distress, minimise harm, and ensure consistent adherence to professional and institutional standards across services.