Application of machine learning algorithms for accurate determination of bilirubin level on in vitro engineered tissue phantom images.

Neonatal Jaundice is a common occurrence in neonates. High excess bilirubin would lead to hyperbilirubinemia, leading to irreversible adverse damage such as kernicterus. Therefore, it is necessary and important to monitor neonates' bilirubin levels in real-time for immediate intervention. However, current screening protocols have their inherent limitations, necessitating more convenient measurements. In this proof-of-concept study, we evaluated the feasibility of using machine learning for the screening of hyperbilirubinemia in neonates from smartphone-acquired photographs. Different machine learning models were compared and evaluated to gain a better understanding of feature selection and model performance in bilirubin determination. An in vitro study was conducted with a bilirubin-containing tissue phantom to identify potential biological and environmental confounding factors. The findings of this study present a systematic characterization of the confounding effect of various factors through separate parametric tests. These tests uncover potential techniques in image pre-processing, highlighting important biological features (light scattering property and skin thickness) and external features (ISO, lighting conditions and white balance), which together contribute to robust model approaches for accurately determining bilirubin concentrations. By obtaining an accuracy of 0.848 in classification and 0.812 in regression, these findings indicate strong potential in aiding in the design of clinical studies using patient-derived images.

Assessment of transient changes in oxygen diffusion of single red blood cells using a microfluidic analytical platform.

Red blood cells (RBCs) capability to deliver oxygen (O2) has been routinely measured by P50. Although this defines the ability of RBCs to carry O2 under equilibrium states, it cannot determine the efficacy of O2 delivery in dynamic blood flow. Here, we developed a microfluidic analytical platform (MAP) that isolates single RBCs for assessing transient changes in their O2 release rate. We found that in vivo (biological) and in vitro (blood storage) aging of RBC could lead to an increase in the O2 release rate, despite a decrease in P50. Rejuvenation of stored RBCs (Day 42), though increased the P50, failed to restore the O2 release rate to basal level (Day 0). The temporal dimension provided at the single-cell level by MAP could shed new insights into the dynamics of O2 delivery in both physiological and pathological conditions.

Vibration motor-integrated low-cost, miniaturized system for rapid quantification of red blood cell aggregation.

Human red blood cells (RBCs) aggregate under low shear conditions, which significantly modulates flow resistance and tissue perfusion. A higher aggregation tendency in blood thus serves as an important clinical indicator for the screening of cardiovascular disorders. Conventional ways of measuring RBC aggregation still require large sample volumes, cumbersome manual procedures, and expensive benchtop systems. These inconvenient and high-cost measurement methods hamper their clinical applicability. Here, we propose a low-cost, miniaturized system to overcome the limitations of these methods. Our system utilizes a coin vibration motor (CVM) to generate a localized vortex for disaggregating RBCs in a disposable fluidic chip. The design of the chip was optimized with fluid dynamics simulations to ensure sufficient shear flow in the localized vortex for RBC disaggregation. The time-dependent increase in light transmittance from an LED light source through the plasma gap while the RBCs re-aggregate is captured with a CMOS camera under stasis conditions to quantify the level of RBC aggregation. Our CVM-based aggregometer was validated against a commercial benchtop system for human blood samples under physiological and pathological conditions, and showed an excellent performance with a high intraclass correlation coefficient of 0.995. In addition, we were able to achieve a rapid measurement (<4 min) with the CVM-based aggregometer, requiring only a 6 µl blood sample. These illustrate the potential of our CVM-based aggregometer for low-cost point-of-care diagnostics without compromising the measurement sensitivity.