In Vitro Evaluation of a Novel Image Processing Device to Estimate Surgical Blood Loss in Suction Canisters

Abstract

Background: Clinicians are tasked with monitoring surgical blood loss. Unfortunately, there is no reliable method available to assure an accurate result. Most blood lost during surgery ends up on surgical sponges and within suction canisters. A novel Food and Drug Administration-cleared device (Triton system; Gauss Surgical, Inc, Los Altos, CA) to measure the amount of blood present on sponges using computer image analysis has been previously described. This study reports on performance of a complementary Food and Drug Administration-cleared device (Triton Canister System; Gauss Surgical, Inc, Los Altos, CA) that uses similar image analysis to measure the amount of blood in suction canisters.

Methods: Known quantities of expired donated whole blood, packed red blood cells, and plasma, in conjunction with various amounts of normal saline, were used to create 207 samples representing a wide range of blood dilutions commonly seen in suction canisters. Each sample was measured by the Triton device under 3 operating room lighting conditions (bright, medium, and dark) meant to represent a reasonable range, resulting in a total of 621 measurements. Using the Bland-Altman method, the measured hemoglobin (Hb) mass in each sample was compared to the results obtained using a standard laboratory assay as a reference value. The analysis was performed separately for samples measured under each lighting condition. It was expected that under each separate lighting condition, the device would measure the various samples within a prespecified clinically significant Hb mass range (±30 g per canister).

Results: The limits of agreement (LOA) between the device and the reference method for dark (bias: 4.7 g [95% confidence interval {CI}, 3.8-5.6 g]; LOA: -8.1 g [95% CI, -9.7 to -6.6 g] to 17.6 g [95% CI, 16.0-19.1 g]), medium (bias: 3.4 g [95% CI, 2.6-4.1 g]; LOA: -7.4 g [95% CI, -8.7 to -6.1 g] to 14.2 g [95% CI, 12.9-15.5 g]), and bright lighting conditions (bias: 4.1 g [95% CI, 3.2-4.9 g]; LOA: -7.6 g [95% CI, -9.0 to -6.2 g] to 15.7 g [95% CI, 14.3-17.1 g]) fell well within the predetermined clinically significant limits of ±30 g. Repeated measurements of the samples under the various lighting conditions were highly correlated with intraclass correlation coefficient of 0.995 (95% CI, 0.993-0.996; P < .001), showing that lighting conditions did not have a significant impact on measurements. Hb mass bias was significantly associated with hemolysis level (Spearman ρ correlation coefficient, -0.137; P = .001) and total canister volume (Spearman ρ correlation coefficient, 0.135; P = .001), but not ambient illuminance.

Conclusions: The Triton Canister System was able to measure the Hb mass reliably with clinically acceptable accuracy in reconstituted blood samples representing a wide range of Hb concentrations, dilutions, hemolysis, and ambient lighting settings.

Figure 1

Triton Canister System uses an iOS app to enable a user to capture scans of blood bearing containers intraoperatively to estimate the Hb mass loss and estimated blood loss within the container

Figure 2

Canister blood volume, total volume, Hb mass, and % hemolysis for prepared samples (n=207)

Figure 3

Scatter plots of Hb mass (g), concentration (g/dl) and EBL (ml) obtained from reference method and Triton System under each of 3 lighting conditions (n=207 for each).

Figure 4

Box plots of Hb mass bias (g) of the same samples under 3 lighting conditions.