Insight

Bridging the Gap Between AI and Clinical Practice in OASI Detection

Obstetric anal sphincter injuries (OASI) are among the most serious forms of birth-related perineal trauma. Around 3 in 100 vaginal births in the UK result in a recorded OASI, rising to around 6 in 100 first vaginal births. Yet recognition and classification at the time of birth remain important challenges. A 2025 NHS Resolution thematic review of OASI-related claims found that 58% of women included in the review had initially had their injury graded as less severe than it actually was, highlighting the potential consequences when sphincter injuries are missed or misclassified.

On paper, careful perineal inspection and digital rectal examination supports accurate identification and classification of perineal trauma. In practice, assessment can be challenging, particularly where anatomy is difficult to interpret and clinicians have differing levels of experience in recognising and classifying perineal trauma.  AI‑enabled tools are now being developed to give an additional, objective signal about sphincter integrity where uncertainty exists.

Why are OASI still missed?

Even with clear guidance, OASI can be difficult to recognise in the moments after birth.

Injuries that are not obvious

Severe perineal trauma is not always easy to see. Oedema, bleeding and tissue trauma can make the extent of injury difficult to interpret, while sphincter damage may not always be immediately apparent on visual examination alone. 

Variations in assessments

Recognising and classifying OASI requires knowledge and experience, and studies have shown variation in detection between examiners. Greater awareness, training and systematic examination can increase the number of injuries identified, meaning higher recorded OASI rates can sometimes reflect better recognition rather than poorer care.

Limited access to EAUS at the point of care

Three‑dimensional endoanal ultrasound (EAUS) is an established reference imaging technique for assessing the anal sphincter complex, but in the UK it is usually available only in specialist perineal or pelvic floor clinics. It is not routinely available as a bedside diagnostic tool immediately after birth. This creates a gap between standard clinical examination and specialist imaging when uncertainty exists over whether the anal sphincter has been injured.

Improving OASI detection with artificial intelligence (AI)

National initiatives such as the OASI Care Bundle have helped strengthen prevention, awareness, and systematic assessment. However, accurate recognition and classification of sphincter injury after birth can still be challenging. This suggests the problem is less about what to do, and more about how reliably it can be done on every shift. It is here that AI‑enabled solutions can help support more accurate detection.

In principle, AI can add value in three ways. It can provide an extra, objective signal at the bedside that is less dependent on the examiner’s level of experience. It can help standardise decisions across different staff and sites, so that a woman or birthing person’s chance of having a sphincter injury recognised does not depend so heavily on who happens to be on duty. Lastly, it can generate structured data that make it easier to audit where injuries are being picked up and where they are being missed.

This technology is not intended to replace perineal inspection or rectal examination but to provide an additional source of information alongside clinical assessment, particularly where there is uncertainty over sphincter integrity. 

What good human–AI collaboration looks like in OASI care

If AI‑enabled tools are going to help rather than hinder OASI detection, they have to be integrated carefully into clinical processes. 

Clinical governance and validation

Introducing any tool should start with shared agreement on when it is used, how results are recorded, and what happens when its output does not match the clinical impression.  Early use should be paired with local validation and ongoing audits. If the algorithm is updated, there needs to be clear version control and a simple way of checking that its behaviour has not changed in ways that matter for safety.

Clinician trust and decision‑making

Trust is unlikely to come from accuracy figures alone. It will grow if clinicians see that the tool behaves predictably. Just as important is explicit reassurance that a “no OASI detected” result does not overrule a strong clinical concern. 

Training and everyday use

Teams should feel confident explaining the examination to women and birthing people, understanding what a possible OASI result actually means, and knowing the agreed next steps when it appears on screen. 

ONIRY: an AI‑enabled tool for bedside OASI detection

ONIRY is an AI‑enabled system that takes a short endoanal measurement and turns it into a simple result about whether sphincter injury is likely. In a prospective multicentre study of 152 women following vaginal birth, ONIRY demonstrated high diagnostic performance compared with EAUS, with sensitivity of 90.6% and specificity of 84.6%. No device-related adverse events were reported.

Used alongside perineal inspection and rectal examination, ONIRY offers teams an extra check when anatomy is difficult or uncertainty is high, helping to make OASI recognition more consistent across different shifts and sites.

To learn more about how ONIRY can support OASI detection in your service, visit the ONIRY product page. You can also contact the Maternity by Kimal team for more information.

ONIRY product page