What is the pouch of Douglas and why does it matter in endometriosis?

The pouch of Douglas is a fold of peritoneum between the uterus and the rectum. In endometriosis, particularly deep infiltrating endometriosis, lesions can infiltrate this space and cause obliteration — where the surfaces adhere. This is a significant finding associated with more advanced disease and is linked to dyspareunia (pain during sex), bowel symptoms, and infertility. It has historically required experienced sonographers or MRI specialists to detect. EndoFusion aims to automate detection from imaging data.

Why does endometriosis take 7 years to diagnose in Australia?

The diagnostic delay reflects several compounding problems: symptom normalisation (many women are told period pain is normal), limited awareness among GPs of the range of endometriosis presentations, scarcity of trained endometriosis surgeons and specialist ultrasound services, and the fact that laparoscopy — historically the gold standard for diagnosis — is a surgical procedure that requires specialist access. Regional and rural women face the longest delays because specialist services are concentrated in major cities. The EndoFusion project specifically targets this access gap.

Can my GP refer me for endometriosis imaging now?

Yes — GP-initiated pelvic ultrasound (transvaginal where appropriate) is a standard first step in the investigation of suspected endometriosis in Australia. Specialist endometriosis ultrasound — particularly for deep infiltrating disease — requires an experienced sonographer and is available at some public and private centres. EndoFusion is not yet clinically available; the current standard of care involves ultrasound ± MRI interpreted by an experienced specialist sonographer, with surgical diagnosis reserved for cases where imaging is inconclusive and clinical suspicion is high.

Australian AI tool detects endometriosis markers with 83% accuracy

Seven years. That is the average time between a woman first experiencing endometriosis symptoms and receiving a diagnosis in Australia. Seven years of normalised pain, dismissed consultations, and delayed access to treatment.

A team at Adelaide University’s Robinson Research Institute has developed an AI tool called EndoFusion that detects two key endometriosis imaging markers from pelvic ultrasound and MRI scans with 83% overall accuracy. The research, published in September 2026, represents one strand of a broader effort to close an access gap that has been a persistent failure of Australian women’s healthcare.

What the tool does

EndoFusion is a multimodal AI system — it processes both ultrasound and MRI data simultaneously and identifies two specific markers of endometriosis:

Obliteration of the pouch of Douglas (POD). The pouch of Douglas is the peritoneal fold between the uterus and the rectum. In deep infiltrating endometriosis, adhesions in this space indicate more advanced disease and are associated with significant symptoms including deep dyspareunia, bowel symptoms, and subfertility. Identifying POD obliteration on imaging has historically required experienced specialist sonographers.

Bowel nodules. Endometriosis lesions infiltrating the bowel wall — another feature of deep infiltrating disease that requires specialist-level interpretation on MRI.

The tool processes scans in milliseconds. Its diagnostic performance reached 83% overall accuracy, with area under the curve (AUC) of 0.765 for POD obliteration on MRI and 0.903 for bowel nodules on MRI. These are early-phase results from a research cohort, not a clinical trial.

The access problem it is trying to solve

The diagnostic delay that EndoFusion targets is not a single-point failure. It is the accumulation of several systemic gaps operating simultaneously.

First: symptom normalisation. Pain with menstruation is common; endometriosis-related dysmenorrhoea is often dismissed as normal, by patients themselves, by families, and by clinicians who do not probe further. The median time from symptom onset to first medical consultation is often years on its own.

Second: GP-level recognition. Endometriosis presents heterogeneously — pelvic pain, dyspareunia, dysmenorrhoea, bladder and bowel symptoms, sub-fertility, or combinations of these — and does not map neatly to a single symptom cluster. The RACGP has updated its endometriosis clinical guidance, but recognition at the GP level remains inconsistent, particularly for presentations without classic cycle-linked pain.

Third: specialist access. Expert endometriosis imaging — specialist transvaginal ultrasound with direct visualisation of deep infiltrating disease, or MRI with dedicated sequences — is concentrated in major metropolitan centres and at a small number of specialist services. For women in regional and remote Australia, access requires a GP referral to a tertiary centre and a wait.

Associate Professor Jodie Avery of the Robinson Research Institute stated the goal directly: “We want to democratise endometriosis by improving access for women in rural and remote communities.” EndoFusion, if validated and deployed, could serve as a plug-in tool for general radiology practices — enabling endometriosis marker detection without specialist sonographer expertise at every site.

Both-and: real promise and early stage

The AI accuracy figures are promising. An AUC of 0.903 for bowel nodule detection on MRI is strong. The 83% overall accuracy is meaningful for a first-generation tool using a multimodal framework.

The caveats matter here. Researchers explicitly note that validation in larger, diverse cohorts is required before clinical deployment. The training cohort, details of which were not fully described in the news coverage, likely came from specialist centres where imaging was already interpreted to a high standard — a setting that may not represent the conditions under which a deployed tool would operate.

AI diagnostic tools in women’s health have a mixed recent history in terms of translation from promising research results to clinical utility. The mechanism here is plausible, the need is genuine, and the performance metrics are early but reasonable. None of that means the tool will be in your radiology software within two years.

What GPs can do now

The practical question for GPs is not “when will EndoFusion be available?” It is: “what is the current best standard for imaging referral in suspected endometriosis?”

The current pathway is:

GP-initiated pelvic ultrasound (transvaginal where appropriate and consented) as first-line imaging
Specialist endometriosis ultrasound by an experienced sonographer where transvaginal findings are inconclusive or where deep infiltrating disease is suspected
MRI with dedicated endometriosis sequences at experienced centres for complex cases
Surgical diagnosis reserved for cases where imaging is inconclusive and clinical suspicion remains high, or where surgical management is planned

The seven-year diagnosis delay is not primarily an imaging technology problem — it is an awareness, recognition, and access problem. Every GP who asks about cyclical pelvic pain, dyspareunia, and bowel symptoms in a menstruating patient, and who does not normalise significant pain without investigation, is operating at the most important point in the pipeline.

Verdict: maybe — watch this space; EndoFusion needs validation, but the problem it targets is real and the diagnostic gap for women is overdue for a solution.

Sources cited

Can AI diagnose endometriosis? — Medical Republic, 24 September 2026: https://www.medicalrepublic.com.au/can-ai-diagnose-endometriosis/129360