نوع مقاله : Original Articles
نویسندگان
1 دکترای تخصصی، گروه بیماریهای دهان، فک و صورت، دانشکدهی دندانپزشکی، دانشگاه آزاد اسلامی، واحد اصفهان (خوراسگان)، اصفهان، ایران.
2 گروه بیماریهای دهان، فک و صورت، دانشکدهی دندانپزشکی، دانشگاه آزاد اسلامی، واحد اصفهان (خوراسگان)، اصفهان، ایران.
3 گروه پاتولوژی، دانشکدهی پزشکی، دانشگاه علوم پزشکی اصفهان، اصفهان، ایران.
4 دکترای تخصصی، دانشکدهی متالوژی، دانشگاه صنعتی اصفهان، اصفهان، ایران.
5 گروه برق و کامپیوتر، دانشگاه صنعتی اصفهان، اصفهان، ایران.
چکیده
عنوان مقاله [English]
نویسندگان [English]
Introduction: Due to high mortality rate and prevalence of cancer, early detection is vital and important. Precancerous oral lesions have the potential to lead to SCC and therefore they should be carefully assessed. Brush cytology is a simple method, which obtains a sample from the epithelium. Computer analyses have an important role in interpretation of pathologic samples.
Materials & Methods: An engineering team designed the software with neural networks, and the algorithm was trained with samples obtained from patients. In the second stage, brush cytology samples were collected from 20 patients with cancer and 20 healthy individuals. From each slide 50 digital images was captured with a camera under a microscope. The images were separately entered into the software program. The results were recorded as healthy and unhealthy. Statistical analyses were performed using Excel program.
Results: The software had 91 errors in a total of 2000 digital images. Comparison of the results provided by the software program and those of the scalpel biopsy of the patients with Fisher's exact test did not reveal any significant differences (p value = 0.004).
Conclusion: Based on the results of this study, the designed software exhibited high specificity and sensitivity.
Key words: Brush cytology, Neural networks, SCC.