A pathology-based machine learning method to assist in epithelial dysplasia diagnosis

Epithelial dysplasia (ED) is one of the most important factors in detecting progression an oral tissue alteration towards carcinoma. Its early detection instrumental avoiding tumor. A major difficulty for ED recognized variability pathologist assessments. This study proposes a new method that leverages expertise to design simple and efficient classification system support dysplastic epithelia. We employ multilayer artificial neural network (MLP-ANN) defining regions epithelium be assessed based on knowledge pathologist. The performance proposed solution was statistically evaluated. implemented MLP-ANN presented average accuracy $$87\%$$ , with much inferior obtained from three trained evaluators. Moreover, led results which are very close those using convolutional (CNN) by transfer learning, 100 times less computational complexity. In conclusion, our show structure combined can lead equivalent more complex structures, routinely used literature.