From Human Mesenchymal Stromal Cells to Osteosarcoma Cells Classification by Deep Learning
Early diagnosis of cancer often allows for a more vast choice of therapy opportunities. After a cancer diagnosis, staging provides essential information about the extent of disease in the body and the expected response to a particular treatment. The leading importance of classifying cancer patients at the early stage into high or low-risk groups has led many research teams, both from the biomedical and bioinformatics field, to study the application of Deep Learning (DL) methods. The ability of DL to detect critical features from complex datasets is a significant achievement in early diagnosis and cell cancer progression. In this paper, we focus the attention on osteosarcoma. Osteosarcoma is one of the primary malignant bone tumors which usually afflicts people in adolescence. Our contribution to the classification of osteosarcoma cells is made as follows: a DL approach is applied to discriminate human Mesenchymal Stromal Cells (MSCs) from osteosarcoma cells and to classify the different cell populations under investigation. Glass slides of differ-ent cell populations were cultured including MSCs, differentiated in healthy bone cells (osteoblasts) and osteosarcoma cells, both single cell populations or mixed. Images of such samples of isolated cells (single-type of mixed) are recorded with traditional optical microscopy. DL is then applied to identify and classify single cells. Proper data augmentation techniques and cross-fold validation are used to appreciate the capabilities of a convolutional neural network to address the cell detection and classification problem. Based on the results obtained on individual cells, and to the versatility and scalability of our DL approach, the next step will be its application to discriminate and classify healthy or cancer tissues to advance digital pathology.