Avaliação de Classificadores como Método de Segmentação de Lesões na Córnea

• Pablo Lima UFPI
• Rodrigo Veras UFPI
• Daniel Leite UFPI
• Walace Tales UFPI
• Francisco Mendes UFPI

Abstract

As well as other areas of the body, the eye is the target of several kinds of diseases that directly affect vision, something essential to the human condition. This work proposes an application method for a Computer Aided Diagnosis system that uses corneal images marked by an ophthalmologist. The method seeks to segment the injured areas using Supervised Machine Learning algorithms, evaluating the performance of three different classifiers: Multi-Layer Perceptron, Random Forest and Support Vector Machine. Among the results obtained, the Random Forest algorithm performed best with Accuracy, Sensitivity and Specificity rates of 93.47%, 94.41% and 93.61%, respectively.

References

Flaxman, S. R. and Bourne, R. (2017). Global causes of blindness and distance vision impairment 1990-2020: a systematic review and meta-analysis.

Pereira, G. H. A. and Centeno, J. A. S. (2013). Utilização de support vector machine para classificação multiclasses de imagens landsat tm.

Russel, S. and Norvig, P. (2013). Inteligência Artificial. Elsevier, 3th edition.

Tapan P. Patel, N Venkatesh Prajna, S. F. N. G. V. L. M. N. L. D. K. H. K. andWoodward., M. A. (2017). Novel image-based analysis for reduction of clinician-dependent variability in measurement of the corneal ulcer size.

Vapnik, V. and Cortes, C. (1995). Support-vector networks. machine learning. n. 20, p273–p297, 1995.