27 May 2020
Purdue University technique images the inner eyelids for measurement.
A project at Purdue University has now demonstrated that the built-in camera of a smartphone can be used as a hyperspectral imager, and in combination with statistical learning techniques can allow Hgb levels to be more easily assessed.
The university's overall mobile application has been christened mHematology, and the findings were reported in Optica.
"Our new mobile health approach paves the way for bedside or remote testing of blood hemoglobin levels for detecting anemia, acute kidney injury and hemorrhages, or for assessing blood disorders such as sickle cell anemia," said Young Kim from Purdue University.
"The Covid-19 pandemic has greatly increased awareness of the need for expanded mobile health and telemedicine services."
The project builds on previous investigations into spectroscopic quantification of Hgb, exploiting hemoglobin's distinct absorption spectrum in the IR and visible ranges. A combination of costly optical components and slow data acquisition rates have to date been barriers to mobile health applications of this approach.
However, it is possible to mathematically reconstruct hyperspectral or multispectral data from images taken by a conventional smartphone camera, according to the Purdue team's published paper, and these computational approaches lay the groundwork for spectral super-resolution (SSR) spectroscopy of blood Hgb.
The project's method involves mathematically reconstructing high-resolution spectra of blood Hgb from color values of the red, green, and blue channels acquired by the built-in camera of a smartphone. The researchers selected the inner eyelid as the site to be imaged, because microvasculature is easily visible there, it is easy to access, and has relatively uniform redness. The inner eyelid is also not affected by skin color, which reduces the need for any patient-specific calibrations.
"A dual-channel hyperspectral imaging system allows us to evaluate the performance of spectroscopic blood Hgb measurements from the eyelids," wrote the team. "Using high-resolution spectral data of the eyelids, we establish a statistical learning framework of SSR for a mobile health application using an affordable smartphone."
Data-driven techniques reduce complexity
To perform a hemoglobin measurement with the new technique, the patient pulls down the inner eyelid to expose the small blood vessels underneath, and the smartphone app then takes pictures. The SSR algorithm extracts the detailed spectral information from the camera’s images, and then another computational algorithm quantifies the blood hemoglobin content by detecting its unique spectral features.
The data acquisition procedure incorporates reference measurements of white reflectance standards, to compensate for the spectral responses of the light source and the camera in the system, according to the team.
In early trials using a standard Android smartphone on a small group of subjects, the mobile health test proved able to provide measurements comparable to traditional blood tests over a wide range of blood hemoglobin values. Further work will now include a separate clinical study via Indiana University to assess use of the technique on oncology patients. Monitoring of nutritional status, anemia, and sickle cell disease in patients in India is also planned.
The team believes that the mHematology application serves as an example of how a data-driven technology can minimize hardware complexity, an important consideration for mobile health applications.
"Our work shows that data-driven and data-centric light-based research can provide new ways to minimize hardware complexity and facilitate mobile health," said Kim. "Combining the built-in sensors available in today’s smartphones with data-centric approaches can quicken the tempo of innovation and research translation in this area."