FAIRE UN DON
EN FR
24 Septembre 2024
Cervical cancer

IARC-developed AI outperforms standard tests in a leap towards global health equity

This year, the theme for World Cancer Research Day is “Innovation in Cancer Research Drives Progress Toward Health Equity”. Scientists from the International Agency for Research on Cancer (IARC), in collaboration with engineers from NSV Incorporated (USA), have developed a revolutionary new artificial intelligence (AI) tool that can accurately detect cervical precancers and cancers in images of the cervix taken during appointments for cervical cancer screening. This innovative system will contribute to greater equity in global health because it has been designed specifically for use in low-resource settings, in contrast to many other AI and digital health solutions.

To train and validate the system’s algorithm, IARC researchers used cervical images collected from nearly 1800 women who screened positive for human papillomavirus (HPV) and had a histopathology diagnosis, including some women living with HIV. The system was then evaluated independently using a different set of images that had not been seen by the algorithm during its training or internal validation. The system performed better than both Pap smear cytology and visual inspection with acetic acid (VIA), the tests currently recommended by the World Health Organization (WHO) to triage women who screen positive for HPV.

The research team took steps to limit biases often encountered in AI development. For example, the new tool incorporates an auto-focusing high-acuity camera with specialized optics, ensuring consistency in images. Previous systems that relied on mobile phone cameras for imagery were hindered by lack of generalization because photographs taken by different models of mobile phone sometimes gave different results. Lack of generalization is a bias that limits the ability of an AI model to interpret data that is different in format from the data that it was trained on.

The IARC-developed system also offers significant advantages for use in low- and middle-income settings because it does not require an Internet connection to interpret the images and it is battery-powered.

The WHO Cervical Cancer Elimination Initiative calls for 70% of women to be screened for cervical cancer using a high-performance test by the age of 35 years, and again by the age of 45 years, and WHO strongly recommends the use of HPV testing as a screening method. HPV testing can be performed on self-collected samples, only needs to be done about once every 10 years, and offers a negative predictive value of close to 100%.

Although the HPV detection test is rapidly becoming the screening test of choice globally, an ideal triage test for HPV-positive women is lacking. This innovative AI tool developed by IARC could fill this gap, and upcoming field validation tests in Zimbabwe will show whether it will provide a reliable and objective solution to triage HPV-positive women and help reduce global inequities in cervical cancer screening.

Watch video

Read more about World Cancer Research Day

Read more about the IARC Cervical Cancer Elimination Initiative (CCEI) Team

 

Publication status

Published in section: Actualité du CIRC

Publication date: 24 Septembre, 2024, 6:05

Direct link: https://d2u7e27yy6nebm.cloudfront.net/fr/news-events/iarc-developed-ai-outperforms-standard-tests-in-a-leap-towards-global-health-equity/

© Copyright International Agency on Research for Cancer 2024

Other news