Title: Fine-tuning an Artificial Intelligence Algorithm for Cervical Cancer Triage in a Diverse US Population: The EASTER-US Study
Background: While HPV testing is increasingly adopted in national screening programs, the efficient triage of HPV-positive women remains a critical challenge due to the low positive predictive value of the test. Standard triage methods like colposcopy are resource-intensive, subjective, and often inaccessible in underserved regions, contributing to significant disparities in cervical cancer outcomes. To address this, the "EASTER Project," funded by the National Cancer Institute (NCI), is being conducted in partnership with the International Agency for Research on Cancer (IARC) to evaluate the nGyn device. Developed by Neo Sense Vector (NSV), this portable, point-of-care colposcope features an integrated artificial intelligence (AI) algorithm previously trained on image datasets from India, Thailand, and Zimbabwe. While IARC is coordinating the concurrent evaluation of this technology in Zimbabwe, the algorithm requires fine-tuning to account for the demographic and anatomical diversity of the US population to ensure equitable performance.
Objectives: The primary objective of the EASTER-US study is to fine-tune and externally validate the AI algorithm’s diagnostic accuracy for detecting cervical intraepithelial neoplasia grade 2 or worse (CIN 2+) among HPV-positive women in the United States. Secondary objectives include assessing the acceptability of the device and measuring trust in AI-assisted diagnostics among both patients and healthcare providers to inform future implementation strategies.
Methods: This prospective, cross-sectional validation study will be conducted across four colposcopy clinics associated with the University of Cincinnati and its Cancer Center. The study aims to recruit 250 women aged 25–59 who have tested positive for high-risk HPV. Following informed consent, participants will undergo standard-of-care colposcopy. During the exam, digital images will be captured using the nGyn device at multiple stages: before and after saline application, after 5% acetic acid application, and after Lugol’s iodine application. All participants will undergo cervical biopsy to establish histopathological ground truth. A sub-study will utilize surveys to evaluate provider workflow integration and patient comfort with the technology.
Results: The study is currently in the pre-enrollment phase. The clinical protocol has been finalized, and regulatory approvals have been secured. Clinical teams at the participating sites have initiated training on the nGyn device and image capture protocols. Furthermore, the research team has met with community partners to ensure recruitment strategies effectively reach underserved populations within the catchment area.
Conclusions/Implications for practice or policy: The nGyn system represents a potential bridge between global health innovation and domestic medical practice. By validating this technology in a US setting alongside concurrent global efforts by IARC, this study aims to provide a scalable, objective triage tool. If successful, this AI-supported approach could significantly increase access to screening, streamline management, and reduce cervical cancer disparities in healthcare deserts where access to specialist care is limited.
Background: While HPV testing is increasingly adopted in national screening programs, the efficient triage of HPV-positive women remains a critical challenge due to the low positive predictive value of the test. Standard triage methods like colposcopy are resource-intensive, subjective, and often inaccessible in underserved regions, contributing to significant disparities in cervical cancer outcomes. To address this, the "EASTER Project," funded by the National Cancer Institute (NCI), is being conducted in partnership with the International Agency for Research on Cancer (IARC) to evaluate the nGyn device. Developed by Neo Sense Vector (NSV), this portable, point-of-care colposcope features an integrated artificial intelligence (AI) algorithm previously trained on image datasets from India, Thailand, and Zimbabwe. While IARC is coordinating the concurrent evaluation of this technology in Zimbabwe, the algorithm requires fine-tuning to account for the demographic and anatomical diversity of the US population to ensure equitable performance.
Objectives: The primary objective of the EASTER-US study is to fine-tune and externally validate the AI algorithm’s diagnostic accuracy for detecting cervical intraepithelial neoplasia grade 2 or worse (CIN 2+) among HPV-positive women in the United States. Secondary objectives include assessing the acceptability of the device and measuring trust in AI-assisted diagnostics among both patients and healthcare providers to inform future implementation strategies.
Methods: This prospective, cross-sectional validation study will be conducted across four colposcopy clinics associated with the University of Cincinnati and its Cancer Center. The study aims to recruit 250 women aged 25–59 who have tested positive for high-risk HPV. Following informed consent, participants will undergo standard-of-care colposcopy. During the exam, digital images will be captured using the nGyn device at multiple stages: before and after saline application, after 5% acetic acid application, and after Lugol’s iodine application. All participants will undergo cervical biopsy to establish histopathological ground truth. A sub-study will utilize surveys to evaluate provider workflow integration and patient comfort with the technology.
Results: The study is currently in the pre-enrollment phase. The clinical protocol has been finalized, and regulatory approvals have been secured. Clinical teams at the participating sites have initiated training on the nGyn device and image capture protocols. Furthermore, the research team has met with community partners to ensure recruitment strategies effectively reach underserved populations within the catchment area.
Conclusions/Implications for practice or policy: The nGyn system represents a potential bridge between global health innovation and domestic medical practice. By validating this technology in a US setting alongside concurrent global efforts by IARC, this study aims to provide a scalable, objective triage tool. If successful, this AI-supported approach could significantly increase access to screening, streamline management, and reduce cervical cancer disparities in healthcare deserts where access to specialist care is limited.
EASTER US Study Flowchart