Seventy-two whole blood samples were investigated to determine the rel
ationship between their spectral data measured in the near infrared (N
IR) wavelength region and haemoglobin content based on laboratory data
determined by a routine standard method as reference. Blood samples w
ere obtained from the Ist Department of Medicine, Imre Haynal Universi
ty of Health Sciences. Donors were selected randomly without respect t
o age, sex, state of health or medical treatment, from apparently heal
thy volunteers as well as from ambulatory and hospitalized patients. N
IR spectra were measured with a SPECTRALYZER 1025 (PMC) computerized s
pectrophotometer in the 1000-2500 nm wavelength region. The relationsh
ip between laboratory data and values of the second derivative (i.e. s
econd order finite difference) of the log(1/TF) spectra measured at di
fferent wavelengths was determined by multiple linear regression (MLR)
using three-and four-term linear summation equations. The cross-valid
ated standard error of performance (SEP) for haemoglobin was 1.348 g d
L(-1) with a three term model acid 1.251 g dL(-1) with a four term mod
el over the range from 5.9 to 20 g dL(-1). This preliminary study indi
cates that NIR measurements can be directly related to haemoglobin con
tent and can be used to determine haemoglobin content in human whole b
lood. (C) 1997 Elsevier Science B.V.