Related Experiment Videos
Computerized morphometric discrimination between normal and tumoral cells in oral smears
Irina-Draga Caruntu1, Monica M Scutariu, Gioconda Dobrescu
1Department of Histology, Faculty of Dentistry, University of Medicine and Pharmacy "Gr.T. Popa" Iassy, Iassy 700115, Romania. dicarunt@mail.dntis.ro
Journal of Cellular and Molecular Medicine
|March 24, 2005
Summary
This study introduces a quantitative classifier using neural networks to detect tumoral cells in oral smears. The method analyzes individual cell morphometrics for accurate oral cancer diagnosis.
Area of Science:
- Biomedical Engineering
- Computational Pathology
- Oral Oncology
Background:
- Oral exfoliative cytology offers a rapid assessment of oral lesions.
- Quantitative analysis of cell features can improve diagnostic accuracy.
Purpose of the Study:
- To develop and validate a quantitative classifier for detecting tumoral cells in oral smears.
- To compare the efficacy of individual morphometric feature analysis versus global smear characterization.
Main Methods:
- Computerized analysis of nuclear and cytoplasmic areas of normal and tumoral oral cells.
- Implementation of a neural network classifier trained on morphometric data.
- Digital image processing using Zeiss KS400 environment.
Main Results:
- The neural network classifier accurately identified tumoral cells in oral smears.
- Classification results consistently matched pathological diagnoses across tested cases.
- Individual morphometric features proved more informative than mean values for smear characterization.
Conclusions:
- A quantitative, neural network-based approach using individual cell morphometrics enhances oral exfoliative cytology.
- This method offers a precise and reliable tool for detecting oral tumoral cells.
- Individual cell feature analysis is superior to global smear analysis for diagnostic purposes.