A new artificial intelligence (AI)-based diagnostic tool for colposcopy examinations may improve the accuracy of diagnosing cancerous and precancerous cervical lesions, including cervical intraepithelial neoplasia (CIN), according to a retrospective validation study presented at the 2023 ASCO Breakthrough meeting in Yokohama, Japan.1 The investigators reported that the tool can also be used to identify appropriate biopsy sites in the cervix.
“Currently, there is no certification system for performing colposcopies in Japan, and the quality and interpretation of these examinations vary. There are not many educational opportunities regarding colposcopy, and there is a lack of specialist physicians who are proficient in colposcopy. We anticipate that this device will be utilized in such settings,” explained lead author Akihiko Ueda, MD, of the Department of Gynecology and Obstetrics at Kyoto University, Japan.
Akihiko Ueda, MD
“Our study aimed to develop an AI-based tool that reproduced the colposcopy examination techniques of specialists, to be used as a diagnostic aid by accurately identifying CIN lesions and guiding tissue sampling locations, independent of examiner skill,” he said.
“Software devices equipped with AI tools are being developed in English. Our team plans to collaborate with clinical researchers in other countries and launch the developed device overseas for use at medical institutions,” Dr. Ueda added.
Study Details
An AI-based lesion-detection model was constructed by annotating abnormal colposcopic findings after acetic acid processing in cervical cancer and CIN3 cases for which histologic diagnoses were confirmed by biopsies. The AI-based model was trained on 60 cases of cervical cancer and CIN3 and validated on 150 cases. The developed lesion-detection model was applied to CIN1 and CIN2 cases, and the diagnostic accuracy of assessment based on these lesions was evaluated.
A validation analysis was performed on colposcopy videos of 8,341 patients who had a colposcopy examination for secondary screening of abnormal cervical cytology or follow-up of CIN between 2013 and 2019. Patients in the study had a median age of 41 years and were diagnosed with 7 early-stage cervical cancer cases, 203 CIN3 cases, 276 CIN2 cases, and 456 CIN1 cases.
The diagnostic accuracy of the CIN1 and CIN2 lesions was evaluated according to the lesion area sensitivity, specificity, area under the curve (AUC), and the number of lesions identified. The model identified severe lesions in CIN3 cases with a sensitivity of 85%, a specificity of 73%, an AUC of 0.89 for lesion area, and an accuracy of 95% for the number of lesions identified. It predicted abnormal colposcopy findings in CIN1 cases with a sensitivity of 87%, a specificity of 70%, an AUC of 0.81 for lesion area, and an accuracy of 97% for the number of lesions identified.
In addition, the model predicted abnormal colposcopy findings in CIN2 cases with a sensitivity of 86%, a specificity of 67%, an AUC of 0.81 for lesion area, and an accuracy of 93% for the number of lesions identified. Furthermore, a heatmap display based on lesion prediction allowed visualization of the area with the highest acetic acid intensity corresponding to the actual biopsy locations.
Next Steps
Dr. Ueda and coauthors are developing software with the AI tool, which is designed to run on a Windows PC.
“Although currently in the experimental stage, physicians can use the AI tool by connecting the device to the colposcope’s camera in the clinical settings. In addition, the software has functions to facilitate the creation of a schematic of colposcopic findings and to save still and moving images. Therefore, the software is expected to be used throughout the entire colposcopy practice,” Dr. Ueda noted.
“Note that since it has not been certified as a medical device, its accuracy must be verified in actual clinical settings before it can be used as a diagnostic aid system,” he said.
Dr. Ueda noted that the accuracy of the application could be improved to predict histopathologic diagnoses. Further study is needed to characterize the relationship between chronologic changes in abnormal colposcopy findings and histopathologic diagnosis.
“The study is currently investigational. In the future, our team plans to conduct clinical trials at multicenter facilities, and we plan to seek approval of the developed AI device as a medical device,” Dr. Ueda stated. “The plan is to release the AI device this fall in Japan.”
Additional Perspective
“Colposcopy plays an important role in cervical cancer screening. This study showed that harnessing the power of artificial intelligence in cancer screening could pave the way for a potentially more effective and improved diagnostic performance in cervical cancer care,” said ASCO Expert Roselle B. De Guzman, MD, FASCO.
Roselle B. De Guzman, MD, FASCO
In the United States, colposcopy is standardized. “The American Society for Colposcopy and Cervical Pathology (ASCCP) offers educational programs including live meetings (eg, postgraduate programs, procedural workshops, and an annual meeting), online CME programs, a continually updated website with focused educational and practice management content, and the Journal of Lower Genital Tract Disease. The ASCCP offers a colposcopy mentorship training program that is accredited by the Accreditation Council for Continuing Medical Education. The goal of ASCCP is to improve clinician competence and performance as well as patient outcomes through educational activities. A primary care provider or gynecologist can perform a colposcopy. Specially trained colposcopists [eg, nurse practitioners, physician assistants, and nurse midwives] can also perform the procedure,” Dr. De Guzman said.
“Colposcopy is generally recommended for an abnormal Papanicolaou smear result, abnormal findings during a pelvic exam, or a positive human papillomavirus test,” she explained.
DISCLOSURE: Dr. Ueda reported no conflicts of interest. Dr. De Guzman has received honoraria from Roche Oncology (Philippines), AstraZeneca, Merck Serono, MSD Oncology, Boehringer Ingelheim, and Zuellig Pharma; has served as a consultant or advisor to Roche, Novartis, Boehringer Ingelheim, AstraZeneca, Zuellig Pharma, Eisai, and MSD Oncology; and has been reimbursed for travel, accommodations, and other expenses from Hospira (Philippines), Roche (Philippines), Merck Sharp & Dohme, Eisai, Boehringer Ingelheim, AstraZeneca, Pfizer, and ZPT Amgen.
REFERENCE
1. Ueda A, Yamaguchi K, Kitamura S, et al: Development of an artificial intelligence-based diagnostic system for the detection of abnormal colposcopic findings. 2023 ASCO Breakthrough. Abstract 82. Presented August 4, 2023.
