Medical Image Classification Based on the Hybrid XGBoost Convolutional Neural Networks

Authors

SENAÏ Besma
Computer Sciences Department, University of science and technology–Mohamed Boudiaf Oran
RAHAL Sidi Ahmed hebri
Computer Sciences Department University of science and technology–Mohamed Boudiaf Oran
KHIAT Salim
Systems Engineering Department Polytechnic National School of Oran Algeria

Synopsis

In recent years, Convolutional neural networks (CNNs) are the most effective models of artificial intelligence. They give good feature extraction, but they contain a large number of layers and parameters, which makes learning difficult. Therefore, their direct applicability to low-resource tasks is not easy. On the other hand, the eXtreme Gradient Boosting (XGBoost) model achieves higher global classification efficiency than other alternative models such as Bagging, Adaboost, SVM and Random Forest. Therefore, this paper first proposes the use of a trained convolutional neural network model for feature extraction from medical images, and then the use of these features by the eXtreme Gradient Boosting (XGBoost) algorithm to build softmax classifiers. This new CNN-XGBoost algorithm takes advantage of both CNN model and XGBoost classifier. The results demonstrate that the novel method used is more efficient than other methods, which confirms the relevance of the proposed approach for medical image classification problems.

ICAECE2023
Published
February 5, 2024