Towards use of electronic health records: cancer classification
Siyu Liao, Jiehao Xiao, Yi Xie, Feng Gu · 2017
The electronic health records (EHR) are generating increasing quantities of data like never before. These extensive amount and large varieties of data provide valuable information and bring opportunities for researchers to study, thus to help clinical practice for clinicians and related decision-making for organizational managers, such as disease diagnosis and healthcare policy making. Cancer diagnosis is very important to patients because finding and treating cancer at an early stage can save lives. In this paper, we choose cancer classification as an example to demonstrate the usage of the EHR data. The data are collected and provided by clinics from private practices in New York City. We apply random forest method for cancer diagnosis. The experimental results of our preliminary work show its effectiveness, 68.5% accuracy in classifying 15 different types of cancers. Moreover, five of them are highly identified and classified, reaching the accuracy of 92.84%.