Pattern recognition analysis to the variation of nasal-pharynx cancer patients' trace element levels in samples of hair, whole blood, and tissue

Abstract

Pattern recognition has been used in this paper to analyze trace element levels in patients diagnosed with nasal-pharynx cancer (NPC) and in healthy control subjects. Trace elements such as Zn, Cu, Fe, Mn, and Mg have been tested in samples of hair and whole blood. Through Mahalanobis Distance Decision analysis, we have achieved good classification effects in whole blood samples: Efficiency for distinguishing patients is 96% and that of healthy controls is 90%. Classification hair samples is inferior to whole blood: Decision accuracy for patients is 58% and healthy controls is 90%. These results are also shown in a nonlinear mapping figure. At the same time, we have also determined 5 trace element levels in 16 other cancer patients' nonneoplastic and cancerous tissue, with no significant difference between them: Decision accuracy of cancerous tissue is 63%, and in nonneoplastic tissue is 50%, hence, we cannot identify them. It can be inferred that there is no idiosyncratic change of trace elements in cancer patients' neoplastic tissue, the change of a cancerous person may occur in the whole body.