An investigation on electronic nose diagnosis of liver cancer

Abstract

Liver cancer is a leading cause of cancer deaths worldwide. However, it is often hard to find liver cancer early because signs and symptoms often do not appear until it is in its later stages. Small liver tumors are hard to be detected on a physical exam. In fact, there are no widely recommended screening tests for liver cancer in people who are not at increased risk. Human exhaled breath contains a mixture of hundreds of volatile organic compounds (VOCs), including some implicit information with diseases. As a result, in recent years, electronic nose has become a widely used tool for the diagnosis of various diseases. In the paper, we evaluated the feasibility of an electronic nose in discriminating liver cancer from healthy controls. Exhaled breaths of liver cancer and healthy controls were first collected. And then, principal components analysis (PCA) was used to achieve the main features. Liner discriminant analysis (LDA), distance discriminant analysis (DDA) and support vector machine (SVM) were respectively applied to construct classifier to analyze exhaled breath. Results showed a satisfactory identification rate of liver cancer subject. SVM could get a higher recognition accuracy than other methods. The recognition accuracy could reach as high as 91.67% based on the single sensor signal analysis. As a result, we believe that electronic nose may become a potential tool for early diagnose liver cancer in future, due to its rapid, simple, cost-less and noninvasive.