Prediction of HB, TW, and LDH in normal and leukemia subjects using PLSR analysis

Abstract

There are numerous applications of Near-Infrared spectroscopy (NIRs) in the medical field. For instance, the application of NIRs can be found in the process of determining a blood parameter which acts as biomarker of specific diseased. This paper proposed a non-invasive leukemia early stage screening technique based on blood hemoglobin (HB), total white cell (TW), and lactate dehydrogenase (LDH). The spectral data was acquired by applying NIRs directly at human fingertip and laboratory data was acquired from standard blood drawing procedure. During PLSR analysis the spectral data was compared with laboratory data for validation. The correlative coefficient of prediction (RP) for normal dataset is 0.9168, 0.8876, and 0.8307 for HB, TW, and LDH respectively. The correlative coefficient of prediction (RP) for leukemia dataset is 0.8588, 0.8868, and 0.8307 for HB, TW, and LDH respectively. The blood parameter with the highest R and smallest difference between RMSEC and RMSEP for both normal and leukemia data sets are HB, TW, and LDH accordingly. The high correlation level between the predicted values and the reference values proves the potential of using NIR spectral information for non-invasive early stage leukemia screening.