Improving Cancers Screening Models Based on Multiple Tumor Markers

Screening of cancers has received considerable attention in developed and developing countries owing to the heavy economic and quality-of-life burden caused by cancers. Several tools based on tumor markers have been developed for cancer risk management. However, each tumor marker test has been developed and validated for a specific category of anticipated cancer. In clinical practice, such tests are widely used in monitoring response to cancer treatment, but there is a lack of evidence supporting their use for screening multiple cancers. In our previous result, cancer screening models based on multiple serum tumor markers were designed using machine learning methods, namely support vector machine (SVM), k-nearest neighbor (KNN), and logistic regression, to improve the screening performance for multiple cancers in a large asymptomatic population. In this research project, we explore to further improve the screening performance of cancer screening models from the following aspects : (1) using other machine learning methods to build cancer screening models, (2) integrating feature selection and machine learning methods to build cancer screening models, and (3) using over-sampling, under-sampling as well as instance selection methods to pre-process collected data, and then integrating feature selection and machine learning methods to build cancer screening models.

Project ID:PB10703-1495
External Project ID:MOST106-2221-E182-071