Aprendizado de máquina simbólico e técnicas fractais para caracterizar rejeição em biópsia miocárdica

This work combines symbolic machine learning and multiscale fractal techniques to generate models that characterize cellular rejection in myocardial biopsies and that can base a diagnosis support system. The models express the knowledge by the features threshold, fractal dimension, lacunarity, number of clusters, spatial percolation and percolation probability, all obtained with myocardial biopsies processing. Models were evaluated and the most significant was the one generated by the C4.5 algorithm for the features spatial percolation and number of clusters. The result is relevant and contributes to the specialized literature since it determines a standard diagnosis protocol. © 2013 Springer.

Keywords

multiscale fractal techniques , myocardial biopsies images , symbolic machine learning , C4.5 algorithm , Diagnosis support systems , Lacunarity , Multiscale fractals , Number of clusters , Percolation probability , Symbolic machine learning , Biomedical engineering , Fractal dimension , Learning systems , Solvents , Biopsy

Language

Portuguese

Citation

5th Latin American Congress on Biomedical Engineering (claib 2011): Sustainable Technologies For the Health of All, Pts 1 and 2. New York: Springer, v. 33, n. 1-2, p. 272-275, 2013.