Related Experiment Videos
Significant reduction in the rate of false-negative cervical smears with neural network-based technology (PAPNET
L G Koss1, M E Sherman, M B Cohen
1Department of Pathology, Montefiore Medical Center and the Albert Einstein College of Medicine, Bronx, NY 10467, USA.
Human Pathology
|October 29, 1997
Summary
A new computer-aided diagnosis instrument identified significant numbers of missed abnormalities in cervical Pap smears, improving detection rates for precancerous lesions and carcinoma. This technology enhances cervical cancer screening accuracy.
Area of Science:
- Gynecology
- Pathology
- Biomedical Engineering
Background:
- False-negative cervical Pap smears contribute to cervical cancer morbidity and mortality.
- Conventional screening and quality control methods have limitations in detecting all abnormalities.
Purpose of the Study:
- To evaluate a novel neural network-based vision instrument for computer-aided diagnosis (CADx) in cervical smear screening.
- To assess the instrument's ability to detect missed cytological abnormalities.
Main Methods:
- Utilized 487 archival negative smears from women with confirmed high-grade lesions or carcinoma (index smears).
- Analyzed 9,666 sequential negative smears as controls.
- Employed a neural network-based vision instrument to identify abnormalities like squamous intraepithelial lesions (SIL) and carcinoma.
Main Results:
- The instrument identified 98 false-negative smears in 72 women (31.6%), including 44 cases of SIL or carcinoma (19.3%).
- SILs were detected in 1.3% of control smears.
- The instrument increased SIL detection by 25% and quality control rescreening yield 5.1-fold compared to historical data.
Conclusions:
- Conventional cervical smear screening and quality control are insufficient for detecting a substantial number of abnormalities.
- The developed CADx instrument significantly improves the accuracy and detection rates in cervical smear screening.
- This technology offers a promising advancement for reducing cervical cancer through more effective screening.