Robert A. Peura

Multivariate Determination of Glucose in Whole Blood Using Partial Least-Squares and Artificial Neural Networks Based on Mid-Infrared Spectroscopy

The infrared (IR) spectra of whole blood EDTA samples, in the range between 1500 and 750 cm−1, obtained from the patient population of a general hospital, were used to compare different multivariate calibration techniques for quantitative glucose determination. Ninety-six spectra of whole undiluted blood samples with glucose concentration ranging between 44 and 291 mg/dL were used to create calibration models based on a combination of partial least-squares (PLS) and artificial neural network (ANN) methods. The prediction capabilities of these calibration models were evaluated by comparing their standard errors of prediction (SEP) with those obtained with the use of PLS and principal component regression (PCR) calibration models in an independent prediction set consisting of 31 blood samples. The optimal model based on the combined PLS-ANN produced smaller SEP values (15.6 mg/dL) compared with those produced with the use of either PLS (21.5 mg/dL) or PCR (24.0 mg/dL) methods. Our results revealed that the combined PLS-ANN models can better approximate the deviations from linearity in the relationship between spectral data and concentration, compared with either PLS or PCR models.

Blood glucose measurement by multiple attenuated total reflection and infrared absorption spectroscopy

The difficulty of measuring physiological concentrations of glucose in blood by conventional infrared absorption spectroscopy is due to the intrinsic high background absorption of water. This limitation can be largely overcome by the use of a CO/sub 2/ laser as an infrared source in combination with a multiple attenuated total reflection (ATR) technique. To demonstrate the applicability of this technique, in vitro measurements of glucose in blood obtained from an experimental infrared laser spectrometer were compared with independent measurements made by a standard YSI 23A laboratory glucose analyzer. The capability of continuous measurement of blood glucose concentration is of primary importance in the future development of a glucose sensor for diabetic patients.<<ETX>>

Algorithm for tissue ischemia estimation based on electrical impedance spectroscopy

The purpose of this paper is to present an algorithm developed for real-time estimation of skeletal muscle ischemia, based on parameters extracted from in vivo obtained electrical impedance spectra. A custom impedance spectrometer was used to acquire data sets: complex impedance spectra measured at 27 frequencies in the range of 100 Hz-1 MHz, and tissue pH. Twenty-nine in vivo animal studies on rabbit anterior tibialis muscle were performed to gather data on the behavior of tissue impedance during ischemia. An artificial neural network (ANN) was used to quantitatively describe the relationship between the parameters of complex tissue impedance spectra and tissue ischemia via pH. The ANN was trained on 1249, and tested on 946 ischemic tissue impedance data sets. A correlation of 94.5% and a standard deviation of 0.15 pH units was achieved between the ANN estimated pH and measured tissue pH values.

Bedside Screening for Venous Thrombosis Using Occlusive Impedance Phlebography

In hospital practice, patients are often seen with signs or symptoms suggesting deep vein thrombosis. This condition is notoriously difficult to diagnose on clinical grounds alone, but it is nevertheless important to establish the diagnosis promptly. Failure to treat the patient early may result in pulmonary embolism or postphlebitic syndrome. On the other hand, anticoagulants should not be prescribed without convincing evidence of deep vein thrombosis since they occasionally cause serious complications due to bleeding. There is an obvious need for a simple method to determine whether

Glucose determination in simulated blood serum solutions by Fourier transform infrared spectroscopy: investigation of spectral interferences

Abstract Determination of physiological concentrations of glucose in whole blood or blood serum using infrared (IR) spectrometry is complicated due to combined effects of spectral variations caused by fluctuations in temperature, pH and other blood constituents with overlapping spectra. In order to initiate systematic examination of these effects, we studied the effects of temperature and pH changes on the spectral variation of phosphate buffered saline (PBS) solutions and glucose doped PBS solutions in vitro. We observed that temperature and pH variations in the glucose doped PBS solutions cause significant changes in absorbance recorded with a Fourier transform infrared/attenuated total reflectance apparatus in the spectral region which contains information about glucose. Primary blood constituents which may interfere with the IR spectrophotometric measurement of glucose in serum were identified. Blood serum solutions were simulated by mixing glucose and the primary interfering constituents in their physiological concentrations with PBS. The feasibility of accurate prediction of physiological glucose concentration in simulated serum solutions covering physiological variations of blood constituents was assessed by applying univariate techniques, multivariate statistical methods and artificial neural networks (ANN) to their mid-IR spectra. Multivariate methods based on partial least squares, principal component regression and ANN produced calibration models with smaller standard errors of prediction (SEP) of 16.9, 18.8 and 18.8 mg dl -1 , respectively, compared with univariate methods based on peak height and area determinations which yielded a smallest SEP of 40.1 mg dl -1 . We conclude that in spite of physiological variations of major interfering constituents, physiological glucose concentration in aqueous multicomponent mixtures such as blood serum may be predicted with sufficient accuracy for clinical applications using multivariate chemometric techniques.

Real-time extraction of tissue impedance model parameters for electrical impedance spectrometer

AbstractThis paper presents a new algorithm for real-time extraction of tissue electrical impedance model parameters from in vivo electrical impedance spectroscopic measurements. This algorithm was developed as a part of a system for muscle tissue ischemia measurements using electrical impedance spectroscopy. An iterative least square fitting method, biased with a priori knowledge of the impedance model was developed. It simultaneously uses both the real and imaginary impedance spectra to calculate tissue parameters R0, R∞, α and τ. The algorithm was tested with simulated data, and during real-time in vivo ischemia experiments. Experimental results were achieved with standard deviations of

Muscle tissue ischemia monitoring using impedance spectroscopy: quantitative results of animal studies

\sigma _{R_0 } = 0.80\% , \sigma _{R_\infty } = 0.84\%

Comparison of Multivariate Calibration Techniques for Mid-IR Absorption Spectrometric Determination of Blood Serum Constituents

, σα=0.72%, and στ=1.26%. On a Pentium II based PC, the algorithm converges to within 0.1% of the results in 17 ms. The results show that the algorithm possesses excellent parameter extraction capabilities, repeatability, speed and noise rejection.