Análise de tecidos normais e anormais de imagens de exames de papanicolau por meio de visão computacional.

Abstract

By using cell images from Pap tests, this project aims to evaluate, with computer perspective aid, if these cells are healthy or cancerous. The main motivation behind the development of this study is the difficulty in identifying through naked eye some particularities of this type of cell, since it is not easy to distinguish one type from another. Dividing these cells into two groups, healthy and cancerous, this work verifies if the system is able to detect and characterize those different kinds of cells. This scientific exposition has used some analysis of texture methods, such as First Order Statistics and Co-occurrence Matrices. The feature vectors acquired from these methods will be then classified by the LDA method (Linear Discriminant Analysis). For the implementation of this project, the software MATLAB was used. By using this software, several experiments were performed in order to demonstrate the feasibility of this project. The final goal was to achieve the highest possible percentage of correct answers, finding a result in the range of 78.63% to hit with the first method, and approximately 80.81% with the latter.