Accuracy of detection of carboxyhemoglobin and methemoglobin in human and bovine blood with an inexpensive, pocket-size infrared scanner

Abstract

Detecting life-threatening common dyshemoglobins such as carboxyhemoglobin (COHb, resulting from carbon monoxide poisoning) or methemoglobin (MetHb, caused by exposure to nitrates) typically requires a laboratory CO-oximeter. Because of cost, these spectrophotometer-based instrument are often inaccessible in resource-poor settings. The aim of this study was to determine if an inexpensive pocket infrared spectrometer and smartphone (SCiO®Pocket Molecular Sensor, Consumer Physics Ltd., Israel) accurately detects COHb and MetHb in single drops of blood. COHb was created by adding carbon monoxide gas to syringes of heparinized blood human or cow blood. In separate syringes, MetHb was produced by addition of sodium nitrite solution. After incubation and mixing, fractional concentrations of COHb or MetHb were measured using a Radiometer ABL-90 Flex® CO-oximeter. Fifty microliters of the sample were then placed on a microscope slide, a cover slip applied and scanned with the SCiO spectrometer. The spectrograms were used to create simple linear models predicting [COHb] or [MetHb] based on spectrogram maxima, minima and isobestic wavelengths. Our model predicted clinically significant carbon monoxide poisoning (COHb ≥15%) with a sensitivity of 93% and specificity of 88% (regression r2 = 0.63, slope P<0.0001), with a mean bias of 0.11% and an RMS error of 21%. Methemoglobinemia severe enough to cause symptoms (>20% MetHb) was detected with a sensitivity of 100% and specificity of 71% (regression r2 = 0.92, slope P<0.001) mean bias 2.7% and RMS error 21%. Although not as precise as a laboratory CO-oximeter, an inexpensive pocket-sized infrared scanner/smartphone detects >15% COHb or >20% MetHb on a single drop of blood with enough accuracy to be useful as an initial clinical screening. The SCiO and similar relatively low cost spectrometers could be developed as inexpensive diagnostic tools for developing countries.

Fig 1

The SCiO (Consumer Physics Ltd., Israel) pocket infrared scanner (Panel A) and schematic diagram of the device in position for scanning a 2.5 cm x 7.5 cm microscope slide with a coverslip and a drop of blood (Panel B).

Fig 2. Detection of carboxyhemoglobin (COHb) in thin films of 50 μL of human and bovine blood in vitro with spectrograms from the SCiO pocket molecular sensor.

A. Representative spectrograms of human blood with varying fractional content of COHb, produced by adding pure CO gas to heparinized blood. Hemoximeter measured COHb is indicated, expressed as % of total hemoglobin, as are the wavelengths of interest in the analysis (775 nm, 875 nm and 925 nm). The multiple traces of the same color indicate repeat scans of the same sample, to indicate reproducibility. B. Regression analysis of spectrometric model [COHb] = f(850nm spectrogram signal-775 nm spectrogram signal)/925 nm signal and hemoximeter measured COHb levels in human and C. bovine blood, based on creation of an index relating differences in spectrogram signals at 875 and 775 nm relative to an isobestic wavelength of 925 nm (see text). D. Modified Bland-Altman bias plot of errors in SCiO estimates of [COHb] % based on the regression equation relating SCiO index value and hemoximeter measured [COHb]. Dashed lines show 95% confidence limits for the regressions.

Fig 3. Detection of methemoglobin (MetHb) in 50 μL thin films of human and bovine blood with in vitro with spectrograms from the SCiO pocket molecular sensor.

A. Representative SCiO spectrograms of human blood with increased fractional content of MetHb, produced by adding NaNO₃ solution to syringes of heparinized blood. The wavelengths used in the spectrophotometric model (775 nm, 875 nm and the isobestic wavelength of 975 nm) are indicated by the vertical lines). B. Regression analysis comparing the spectrometric model and hemoximeter measured MetHb levels in human blood. C. Regression analysis comparing the spectrometric model and hemoximeter measured MetHb levels in bovine blood. D. Modified Bland-Altman bias plot of errors in SCiO estimates of [MetHb] % based on the regression equations relating SCiO MetHb Index value and hemoximeter measured [MetHb] for human and bovine blood.