Machine Learning Approach to Classify and Predict Osteosarcoma Grading

Osteosarcoma is a type of bone cancer. The malignant tumors are mainly found in cells that forms the bone. Osteosarcoma grows in long bones near metaphyseal growth plates, commonly found in Humerus (10%), Tibia (19%), and femur (42%). ^[1] To identify a malignant bone tumor is a strenuous task as it is difficult to find it Also, classification from histopathological images of Osteosarcoma is a cumbersome and tedious process. Generally, Osteosarcoma is classified into viable, Non-viable, and Non-tumor classes.In this paper, required features were extracted from the histopathological images, which were obtained from The Cancer Imaging Archive, using CellProfiler and the features were then fed to the four machine learning models used in this project. Random Forest (RF), Support Vector Machine (SVM) Decision Trees (DT) and Logistic Regression (LR) are the machine learning algorithms used. These models were used to achieve a more accurate classification of Osteosarcoma into its respective three classes. Finally, the outcomes of all the models were compared to select the best performing of them.