Predicting Osteosarcoma using eXtreme Gradient Boosting Model

Bone cancer is when unusual cells grow out of control in our bone. It destroys normal bone tissue. Osteosarcoma, also called osteogenic sarcoma, is a kind of bone cancer. It happens when the new bone tissue cells form a cancerous tumor. These cancerous cells seem to be nascent stages of bone cells. However, what differentiates them from normal bone cells is that the bone tissue in Osteosarcoma is not as strong as that in normal bones. It generally occurs in adolescent children. Osteosarcoma can be identified by image tests of the bone such as X-rays, Computerized Tomography (CT) scans, and Magnetic resonance imaging (MRI) scans. It can also be identified by conducting a biopsy. This paper proposes a novel way to classify different Osteosarcoma types with a high accuracy using histology images (Microscopic bone images). Analysis of bone using histology is a tedious and lengthy process. In this paper, the eXtreme Gradient Boosting (XGBoost) algorithm is used to classify Osteosarcoma achieving an accuracy of 94.84 %, a sensitivity of 94.83% and a specificity of 97.41 %. The proposed model will aid the physicians to drastically speed up the diagnosis process.