- Politics
- Diversity, equity and inclusion
- Financial Decision Making
- Telehealth
- Patient Experience
- Leadership
- Point of Care Tools
- Product Solutions
- Management
- Technology
- Healthcare Transformation
- Data + Technology
- Safer Hospitals
- Business
- Providers in Practice
- Mergers and Acquisitions
- AI & Data Analytics
- Cybersecurity
- Interoperability & EHRs
- Medical Devices
- Pop Health Tech
- Precision Medicine
- Virtual Care
- Health equity
Machine Learning Detects Atypical Cervical Smears
The tech is feasible to use in resource-limited setting to detect abnormalities and under-diagnosed diseases.
Johan Lundin, M.D., Ph.D.
Digital microscopy with artificial intelligence (AI) helps detect atypical cervical smears with a high sensitivity compared with visual sample analysis.
The findings suggest the technologies are feasible to use in rural, resource-limited settings for the detection of abnormal cells in Papanicolaou tests.
Johan Lundin, M.D., Ph.D., and colleagues determined whether AI-supported digital microscopy diagnostics could be implemented in such settings to analyze such tests because inadequate access to microscopy diagnostics is a problem in limited-resource areas and impairs the diagnosis of common and treatable conditions. The research site was located in rural Kenya.
The team acquired Papanicolaou smears from 740 women attending a regional HIV-control program between September 1, 2018 and September 30, 2019, from patient volunteers who were not pregnant, aged between 18 and 64 years old, had confirmed HIV positivity, and signed informed consent.
Patients were assigned a study number and trained nurses obtained the Papanicolaou tests. The tests were fixed and stained with the Papanicolaou staining method. Staining quality was evaluated by light microscopy and the slides were then digitized, stored in slide boxes, and transported to the pathology laboratory. Digitized and physical slides were pseudonymized using study numbers. Papanicolaou slides were digitized with a portable whole-slide microscope scanner and deployed in a laboratory.
The investigators used a commercially available machine-learning and image-analysis platform. They trained an algorithm based on deep convolutional neural networks to detect low-grade squamous intraepithelial lesions (LSIL) and high-grade squamous intraepithelial lesions (HSIL) in the smear digital whole slides. The samples were split into a training and tuning set (n=350) and a validation set (n=390).
A trained pathologist analyzed physical slides with light microscopy. Adequate slides were reviewed according to the Bethesda classification system. Slides with findings recorded as LSIL or higher were included as slides with significant cervical cell atypia.
The mean age of the women included in the study was 41.8 years old. For the detection of cervical cellular atypia, sensitivities were 95.7% (95% CI, 85.5-99.5) and 100% (95% CI, 82.4-100) and specificities were 84.7% (95% CI, 80.2-88.5) and 78.4% (95% CI, 73.6-82.4), in comparison with the pathologist assessment of digital and physical slides, respectively. The areas under the receiver operating characteristic curve were .94 and .96. The investigators noted the negative predictive values were high (99%-100%) and accuracy was high, mostly for the detection of high-grade lesions. The AI did not have any false-negatives for samples classified as high grade by manual sample analysis.
The technology could create new opportunities to facilitate the diagnostics of a variety of diseases that are still under-diagnosed, especially in low-resource settings, Lundin and the team concluded.
The study, “Point-of-Care Digital Cytology With Artificial Intelligence for Cervical Cancer Screening in a Resource-Limited Setting,” was published online in JAMA Network Open.
