By Chiara Tagliavini
Ultrasound has become an increasingly popular imaging modality in medical practice worldwide. It is especially valued for its safety, as it uses non-ionising sound waves and has shown no adverse biological effects at diagnostic levels1.
A growing area of innovation is the integration of artificial intelligence (AI) into ultrasound imaging. AI, particularly deep learning, shows strong potential in medical diagnostics by recognising complex patterns in visual data, a natural complement to sonography2.
In gynaecology, development of AI applications is already gathering pace and there’s one disease area where it shows great promise: endometriosis. The disease affects around one in ten women of reproductive age, yet it remains one of the most underdiagnosed and misunderstood gynaecological conditions3. According to Endometriosis UK, many women wait seven to eight years on average for a diagnosis, often enduring chronic pain, infertility, and repeated medical visits. This delay not only impacts quality of life but also increases the risk of disease progression. Faster, more accessible tools are urgently needed to help close this diagnosis gap.
Dr Parisa Ensafi, MD: “Endometriosis is far more common than it appears. Yet, it often goes undiagnosed because the most reliable diagnostic method, laparoscopy, is an invasive surgical procedure requiring general anesthesia and comes with significant costs. If a widely accessible, non-invasive tool like ultrasound could reliably detect endometriosis, it would mark a revolutionary breakthrough in gynecological care.”
AI applications for endometriosis diagnosis are relatively new but rapidly expanding. A 2022 review found that half of all relevant studies had been published in the previous five years. While current ultrasound-AI methods are still in early stages, initial studies report promising results, with sensitivities of up to 88% and specificities near 90%3. The market for AI in medical imaging was projected to exceed $2 billion by 2023.
Ultrasound, particularly transvaginal ultrasound (TVUS), is a key non-invasive tool for assessing pelvic pathology. Unlike laparoscopy, the current standard for diagnosing endometriosis, TVUS carries no surgical risk and is more accessible. It is widely available, cost-effective, and increasingly portable, making it useful even in low-resource or remote settings4.
Moreover, ultrasound provides real-time imaging, enabling immediate correlation with symptoms, a major advantage in both routine assessments and emergency gynaecological care.
Despite these strengths, TVUS has limitations in diagnosing endometriosis, especially deep infiltrating endometriosis (DIE). Subtle findings can be easily missed, and accurate interpretation depends heavily on the operator’s experience3. This diagnostic variability highlights where AI can make a meaningful difference.
Yet, while TVUS is common in clinical settings, few AI studies have focused on ultrasound- based diagnosis of endometriosis. A 2022 review identified only three studies that incorporated imaging-derived features as primary AI inputs. One of the features that stood out is the absence of the sliding sign, an important ultrasound marker used to assess pelvic involvement. This feature appeared consistently across all three studies and seems especially suited for AI tools, since it’s a clear visual sign that carries real diagnostic value3.
Ultrasound’s real-time nature makes it especially well-suited for AI integration. Having a system that flags features like kissing ovaries or an absent sliding sign during a live scan, would provide instant diagnostic support. Such tools could reduce delays, improve consistency, and assist clinicians with varying levels of experience.
AI has the potential to standardise interpretation, reduce human error, and support earlier, more accurate diagnoses, particularly for a condition as complex and under-recognised as endometriosis2. Earlier detection means better care, improved treatment planning, and enhanced quality of life.
Though still in its early stages, AI-enhanced ultrasound presents a promising path toward more equitable and effective diagnostic practices in gynaecology. To fully realise this potential, continued investment, interdisciplinary research, and close collaboration across healthcare and technology sectors are essential.
Ultimately, the integration of AI into ultrasound imaging holds the promise of transforming endometriosis care, empowering clinicians with sharper diagnostic tools and offering patients the hope of earlier answers and improved outcomes. As technology and medicine converge, the vision is clear: a future where timely, accurate diagnosis is the norm, and no one is left waiting in the dark.
—————————————————————————————-
Chiara Tagliavini is a medical physics graduate currently pursuing her master’s degree, with research experience in radiobiology at the German Cancer Research Center (DKFZ). She is deeply committed to enhancing patient outcomes through clinical practice and has a strong interest in cancer therapy and its role in modern healthcare.
References
1 Tole, N. M. (n.d.). Basic physics of ultrasonographic imaging. In H. Ostensen (Ed.), Diagnostic Imaging and Laboratory Technology, Essential Health Technologies, Health Technology and Pharmaceuticals. World Health Organization.
2 Drukker, L., Noble, J. A., C Papageorghiou, A. T. (2020). Introduction to artificial intelligence in ultrasound imaging in obstetrics and gynecology. Ultrasound in Obstetrics C Gynecology, 56(4), 498–505. https://doi.org/10.1002/uog.22122
3 Sivajohan, B., Elgendi, M., Menon, C. et al. Clinical use of artificial intelligence in endometriosis: a scoping review. npj Digit. Med. 5, 109 (2022). https://doi.org/10.1038/s41746-022-00638-1
4 Recker, F., Dietrich, C. F., Weber, E., Strizek, B., Gembruch, U., C Campbell Westerway, S. (2021). Point-of-care ultrasound in obstetrics and gynecology. Archives of Gynecology and Obstetrics, 303(4), 871–876. https://doi.org/10.1007/s00404-021-05972-5
