AI-assisted colposcopy to support clinicians detecting precancerous lesions

Why it matters:

Cervical cancer is largely preventable, yet it still claimed almost 350,000 lives worldwide in 2022. Nearly 94% of those deaths occurred in low- and middle-income countries. In Europe, about 58,000 comen develop the disease each year and 27,000 die from it. The WHO elimination strategy aims for 70% of women to be screened with a high-performance test and 90% of women with precancer or cancer to receive treatment by 2030. Colposcopy is the key technique for triaging women with a postitve test result in screening via cytology or HPV-test. Its quality depends on examiner experience, and trained colposcopists are scarce in many regions.

Colpo AI aims to provide standardised, reproducible decision support at these steps.

Our goals:

• Train and refin our AI model
• Build a high-quality clinical data pipeline together with local gynaecologists as clinical partners
• Support clinical decision-making

What we do:

AI-assisted colposcopy: Our algorithm analyses colposcopy images and offers five functions:

• Quality control: It checks image and examination quality. It flags unsuitable images, so that only reliable images feed the analysis.
• Transformation zone classification: it classifies the transformation type.
• Lesion segmentation: it delineates suspicious areas and precancerous lesions in the image.
• Lesion classification: it classifies the segmented findings, to separate normal from abnormal tissue and indicate severity.
• Biopsy spot suggestions: it highlights the areas most suspicious for precancerous lesions, to guide where to take biopsies.

AI-assisted biopsy analysis: A separate algorithm supports the analysis of biopsy histology. This links the colposcopic finding with the pathology result.

Integrated in OncoPath: A prototype of Colpo AI is already integrated into OncoPath, our cloud-based platform that covers the screening pathway from HPV testing and recall to colposcopy, histopathology, treatment and follow-up. Images, histology and clinical data are harmonised across sites. This gives the AI a consistent, well-documented data foundation and gives us an idea how an AI could be integrated into the clinical workflow in the future.

Clinical partnership and validation: We work with gynaecologists and pathologists to ensure clinical relevance and representative data.

Responsible AI: Fairness, transparency, data protection and human oversight are built into every step.

Join us

Are you a gynaecologist, pathologist or research institution interested in
collaborating? Get in touch.