Authors:
Thales Lopes
1
;
Guilherme Roberto
2
;
Carlos Soares
2
;
Thaína Tosta
3
;
Adriano Silva
4
;
Adriano Loyola
5
;
Sérgio Cardoso
5
;
Paulo R. de Faria
6
;
Marcelo Z. do Nascimento
4
and
Leandro Neves
1
Affiliations:
1
Department of Computer Science and Statistics, São Paulo State University, São José do Rio Preto-SP, Brazil
;
2
Faculty of Engineering, University of Porto, Porto, Portugal
;
3
Institute of Science and Technology, Federal University of São Paulo, São José dos Campos-SP, Brazil
;
4
Faculty of Computer Science, Federal University of Uberlândia, Uberlândia-MG, Brazil
;
5
Area of Oral Pathology, School of Dentistry, Federal University of Uberlândia, Uberlândia-MG, Brazil
;
6
Department of Histology and Morphology, Institute of Biomedical Science, Federal University of Uberlândia, Uberlândia-MG, Brazil
Keyword(s):
Deep Learning, Fractal Features, Explainable Artificial Intelligence, Histological Images.
Abstract:
In this work, a method based on the use of explainable artificial intelligence techniques with multiscale and multidimensional fractal techniques is presented in order to investigate histological images stained with Hematoxylin-Eosin. The CNN GoogLeNet neural activation patterns were explored, obtained from the gradient-weighted class activation mapping and locally-interpretable model-agnostic explanation techniques. The feature vectors were generated with multiscale and multidimensional fractal techniques, specifically fractal dimension, lacunarity and percolation. The features were evaluated by ranking each entry, using the ReliefF algorithm. The discriminative power of each solution was defined via classifiers with different heuristics. The best results were obtained from LIME, with a significant increase in accuracy and AUC rates when compared to those provided by GoogLeNet. The details presented here can contribute to the development of models aimed at the classification of hist
ological images.
(More)