Publication: Deep learning feature extraction and classification for osteosarcoma
Loading...
Date
2023-07
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
Osteosarcoma, despite being a rare form of bone cancer, is the most common primary bone malignancy, which primarily affects children and adolescents. Although as technology become more advance which has improved the diagnostic methods, histopathology still stands as one of the good ways for disease staging and treatment decisions. With recent rise in the semiconductor industries and computing power, medical AI and deep learning has shown to be a potential tool to help in analysing and evaluating the histopathology, specifically in classifying the existence and type of tumor. Histopathology images alongside the labels from this study are taken from online medical image archive and are trained using Keras pretrained networks. Ensemble learning method is carried out as a follow experiment to attempt to further improve the accuracy. In result, the VGG16 model is the highest among the network used with a 93.85% overall accuracy on test dataset. By using ensemble model, the accuracy did not improve, with only an accuracy of 91.80% on test dataset. Keywords: osteosarcoma; deep learning; pretrained network; ensemble learning