Fig 1.
Overview of the paper-based SCA diagnostic test.
(a) To perform the test, a 20 μL droplet of blood mixed 1:10 (by volume) with Hb solubility buffer–a concentrated phosphate buffer (2.49M) containing a hemolytic (saponin) and a reducing agent (sodium hydrosulfite)–was deposited on paper. Differential transport of polymerized HbS and soluble forms of Hb in the paper substrate produced blood stains with characteristic patterns. (b) Representative images of the blood stain patterns produced by HbAA, HbAS and HbSS samples.
Fig 2.
Validation of the paper-based SCA test in a research laboratory and in a resource-limited clinical setting (Cabinda, Angola).
(a) Aggregate confusion matrix for the diagnoses of blood samples collected from normal volunteers and patients of Pediatric Hematology-Oncology Clinic at Tulane University Hospital and of the Sickle Cell Center of Southern Louisiana (New Orleans, LA) performed via visual evaluation of the blood stains by human scorers (n = 5). Rows correspond to true genotypes (diagnosed by hemoglobin electrophoresis) and columns correspond to predicted genotypes (diagnosed by the paper-based test). Shaded cells along the diagonal contain numbers of correct diagnoses. (b) Confusion matrix for the diagnoses of blood samples collected at the Primero de Maio obstetric hospital from postnatal females with unknown SCA status. The rapid test was performed and interpreted via visual evaluation by healthcare workers at the newborn screening laboratory of the Clinica de Celulas Falciformes at the Dispensario Materno Infantil (Cabinda, Angola). Rows correspond to true genotypes (diagnosed by isoelectric focusing) and columns correspond to predicted genotypes (diagnosed by the paper-based test). Shaded cells along the diagonal contain numbers of correct diagnoses.
Table 1.
Visual diagnosis performance metrics.
Fig 3.
Automated analysis of blood stains in paper.
(a) A custom image analysis algorithm automatically detected the center of each blood stain (dashed crosshair) and extracted the RGB values for all pixels contained within the dark red center spot (smaller dashed circle) and within the pink peripheral ring (area between the smaller and the larger dashed circles). The S-index was defined as the quotient of the mean red color intensity of the center spot and that of the peripheral ring of the blood stain. (b) The values of the S-index for (●) HbAA (n = 18), (♦) HbAS (n = 17) and (■) HbSS (n = 20) samples obtained from healthy subjects and patients in New Orleans, LA. (c) Receiver operating characteristic (ROC) curves for the use of S-index to identify HbAA, HbAS and HbSS samples. The area under the curve (AUC) for discriminating HbAA from HbAS and HbSS was 1.00, the AUC for discriminating HbSS from HbAA and HbAS was 0.9986 and the AUC for discriminating HbSS from HbAS was 0.9971.
Table 2.
Automated diagnosis performance metrics.
Fig 4.
Diagnosis of other forms of sickle cell disease.
(a) Representative images of blood stains produced in paper by HbSC samples and by HbAS samples, for comparison. (b) Representative images of blood stains produced in paper by HbSβ+-thalassemia samples and by HbSS samples, for comparison. (c) Classification of HbSC and HbSβ+-thalassemia samples in the S-index domain. The values of the S-index for (○) HbAA, HbAS and HbSS samples (n = 55), (■) HbSC (n = 16) and (♦) HbSβ+-thalassemia (n = 3) are shown on Y-axis. The location along the X-axis of each data point for HbAS, HbSS, HbSC and HbSβ+-thalassemia samples corresponds to their %HbS measured with Hb electrophoresis by an off-site clinical laboratory. HbAA samples (n = 18) contained no HbS; hence the random lateral spread of the data points representing these samples on the plot.
Fig 5.
Differentiation of HbSC samples from HbAS samples via automated image analysis.
(a) The values of S-index for HbAS (n = 17) and HbSC (n = 16) samples. (b) Vertical scatter plots of C-index (normalized by cutoff values) for HbAS and HbSC samples. C-index was defined as the product of total color intensity of all pixels in the center area of the blood stain and average color intensity of all pixels at a distance of 5 mm from the center of the blood stain. (c) ROC curves for differentiating HbSC and HbAS based on (-) S-index, AUC = 0.67; (-) C-index, AUC = 0.82; and (-) %HbS by electrophoresis, AUC = 0.99.