Investigation of Noninvasive Near-Infrared Spectroscopic Sensor for Measurement of Blood Hematocrit
AuthorsTran, Anh T.
Faculty AdvisorSoller, Babs R.
UMass Chan AffiliationsDepartment of Anesthesiology
Health Services Research
Medicine and Health Sciences
MetadataShow full item record
AbstractBackground:Hematocrit, the volume percent red cells in blood, is monitored frequently in surgical patients who undergo procedures with large blood losses and in patients suffering from hematological or oncological diseases. The current standard method for measuring hematocrit is not only invasive but also inefficient with a time lag between sampling of the blood and availability of the results. To provide a noninvasive and direct method of hematocrit measurement, our laboratory has developed a near-infrared spectroscopic (NIRS) monitor, a prototype that also provides measurements for muscle oxygen and muscle pH. Past studies have demonstrated the technique to have acceptable accuracy of measuring the mentioned parameters for individual patients having cardiac surgery on cardiopulmonary bypass. (Soller et al. Crit Care Med 2003, 31:9, 2324-31; Zhang et al. Appl Spectrosc 2000, 54:2, 294-9) However, formulas to adjust for subject-to-subject variation are necessary for accuracy across all patients. Recently, the methodology has been refined with an orthogonalization technique that provides simultaneous correction for spectral interference from skin pigment and fat thickness using a two-distance fiber-optic probe. The correction has shown enhanced accuracy of NIRS-based measurements of blood chemistry. (Yang et al. Optics Letters 2005, 30:17) Our hypothesis for this study is that application of the orthogonalization technique in the computation of capillary hematocrit values in the muscle by the NIRS monitor will improve measurement accuracy across several human subjects. Objectives:To contribute to the development of a hematocrit calibration equation for the NIRS monitor and to assess the accuracy of the orthogonalization technique in hematocrit calculations across several human subjects. Methods:Evaluation of the NIRS monitor was conducted at the Hematology/Oncology clinic to obtain a wide range of hematocrit values and in our laboratory with healthy volunteers. In the clinic, ill patients presenting for blood drawn for complete blood count (CBC) were recruited and consented. Hematocrit values from CBC were used as reference hematocrit for this study. The basic procedure involved having the noninvasive NIRS sensor placed on the patient’s forearm for a total of 5 minutes of spectral data collection. The average collected data for the last 2 minutes were utilized to compute a hematocrit calibration equation with and without correction for skin pigment and fat thickness. Spectra, orthogonalized and non-orthogonalized, were used with the hematocrit reference values to develop 2 calibration equations. The calibration equation derived from the orthogonalized spectra was then validated on a set of spectra and blood values from healthy volunteers. Data collection and blood work on these volunteers were done as outlined above. The levels of measurement agreement between the NIRS-based hematocrit values and those obtained from traditional blood work were analyzed with the Bland and Altman method. Results:Data analysis was performed for 17 patients from the Hematology/Oncology clinic and for 9 healthy volunteers. Reference hematocrit ranges for the patients and healthy volunteers are 31% to 42% and 36% to 43%, respectively. This study shows that spectra processed with the orthogonalization technique provide better correlation between the reference and NIRS based hematocrit values comparing to non-orthogonalized result. The linear correlation between orthogonalized spectral absorbance and reference hematocrit values demonstrates an R2 value of 0.651 while for non-orthogonalized spectra, R2 was only 0.003. Thus, a hematocrit calibration equation could only be developed with correction for skin pigment and fat thickness. Validation of the calibration equation on healthy volunteers indicates that the linear correlation between the reference and NIRS calculated hematocrit has an R2 of 0.56 with an estimated measurement error of 2.4 hematocrit units. A Bland and Altman analysis reveals an average bias of 1.4 hematocrit units, and the estimate of error (2 standard deviation) is ± 3.7 hematocrit units. Conclusion:These results demonstrate that by correcting for optical interference of skin pigment and fat thickness with the orthogonalization technique, the noninvasive NIRS monitor provides strong correlation between the NIRS-based hematocrit values and those from the invasive method. The potential of this methodology to measure hematocrit noninvasively, continuously, and in real-time does not only benefit patients requiring frequent hematocrit evaluation but also the pediatric and elderly population. Further study is necessary to expand the hematocrit calibration equation for lower and higher hematocrit values than the ones that we obtained.
Permanent Link to this Itemhttp://hdl.handle.net/20.500.14038/49171
The author participated in this study as a medical student in the Senior Scholars research program at the University of Massachusetts Medical School.