AI-powered Diagnostic Imaging Tool Estimates Hemoglobin Levels

A collaboration between SigTuple Technologies and the Indian Institute of Science demonstrated an AI-powered, imaging-based tool for the estimation of hemoglobin levels. The setup combines a microfluidic chip and an AI-enabled microscope designed for deriving the total, as well as differential counts, of blood cells.

Existing hematology analyzers and many medical devices for multiparameter assessment often have dedicated subcompartments with separate optical detection systems. This design tends to require greater sample volumes and can drive up the cost of the equipment as a whole. The new method instead uses a microfluidic chip, which, including the reagent, costs less than $0.14, according to the research paper, as well as a detector in the form of a conventional microscopy camera.

A diagnostic solution to enable simple and cost-effective hemoglobin estimation using a microfluidic chip and AI-powered automated microscope. Courtesy of SigTuple Technologies Pvt. Ltd.

“In this study, we demonstrate that the applicability of a system originally designed for the purposes of imaging can be extended toward the performance of biochemical tests without any additional modifications to the hardware unit, thereby retraining the cost and laboratory footprint of the original device,” said Srinivasan Kandaswamy, second author on the study.

The device is enabled by the novel design behind the microfluidic chip, a customized biochemical reagent, optimized imaging, and an image analysis procedure specifically tailored to enable the clinical performance of the diagnostic test.

Tests showed that the system obtained data that were comparable to those of the predictions of traditional hematology analyzers, showing a Pearson correlation of 0.99, with a turnaround time of 10 min, including incubation. The method achieved a sensitivity of 92.3% and a specificity of 53.8%.

The validation study demonstrated that the technique meets regulatory standards, meaning that it is likely to be accepted for clinical use by doctors and hospitals.

The automated microscope ordinarily uses a combination of red, blue, and green LEDs; however, in this case, it used only the green light, as the optimized reagent (sodium dodecyl sulfate with hemoglobin) complex absorbs light in the green wavelength.

“This paper lays the foundation and will also serve as a guide to future attempts to translate conventional biochemical assays onto a chip, from point of view of both chip design and reagent development,” Kandaswamy said.

Aside from measuring hemoglobin in the blood, a similar setup with minor modifications could be used to measure protein content, cholesterol, and glycated hemoglobin.

The research was published in AIP Advances (www.doi.org/10.1063/5.0036446).