Osteosarcoma Classification using Convolutional Neural Network
Talitha Asmaria, Dita Ayu Mayasari, M. Ary Heryanto, Menik Dwi Kurniatie, Rahma Wati, Salsabila Aurellia · 2021
A biopsy is the final procedure of the diagnosis process and functions to distinguish the primary and secondary osteosarcomas (OS). The problem comes when many pathologists lead to misdiagnosis in the differentiation of various specimens of neoplasms due to its morphological diversity. This study aims to classify the cell viability of the osteosarcoma’s dataset with hematoxylin and eosin (H&E) stained. To increase efficiency and precision, Matlab is used to construct a convolutional neural network (CNN). The CNN architecture included six convolution layers, max-pooling layers, and fully connected layers for feature extraction. Data augmentation is used to boost performance. The classification process using the CNN architecture produces an average accuracy of 95.37 percent. The accuracy of each class, namely non-tumor is 92.8 percent, non-viable is 98 percent, and viable is 95.239 percent. To conclude, the CNN technique for classifying the OS works in high accuracy numbers. This study can significantly obtain better results compared with the manual process and effectively help the pathologist works.