Handheld multispectral imager for quantitative skin assessment in low-resource settings

Abstract

Significance: Spatial frequency domain imaging (SFDI) is a quantitative imaging method to measure absorption and scattering of tissue, from which several chromophore concentrations (e.g., oxy-/deoxy-/meth-hemoglobin, melanin, and carotenoids) can be calculated. Employing a method to extract additional spectral bands from RGB components (that we named crosschannels), we designed a handheld SFDI device to account for these pigments, using low-cost, consumer-grade components for its implementation and characterization. Aim: With only three broad spectral bands (red, green, blue, or RGB), consumer-grade devices are often too limited. We present a methodology to increase the number of spectral bands in SFDI devices that use RGB components without hardware modification. Approach: We developed a compact low-cost RGB spectral imager using a color CMOS camera and LED-based mini projector. The components' spectral properties were characterized and additional cross-channel bands were calculated. An alternative characterization procedure was also developed that makes use of low-cost equipment, and its results were compared. The device performance was evaluated by measurements on tissue-simulating optical phantoms and in-vivo tissue. The measurements were compared with another quantitative spectroscopy method: spatial frequency domain spectroscopy (SFDS). Results: Out of six possible cross-channel bands, two were evaluated to be suitable for our application and were fully characterized (520 AE 20 nm; 556 AE 18 nm). The other four crosschannels presented a too low signal-to-noise ratio for this implementation. In estimating the optical properties of optical phantoms, the SFDI data have a strong linear correlation with the SFDS data (R 2 ¼ 0.987, RMSE ¼ 0.006 for μ a , R 2 ¼ 0.994, RMSE ¼ 0.078 for μ 0 s). Conclusions: We extracted two additional spectral bands from a commercial RGB system at no cost. There was good agreement between our device and the research-grade SFDS system. The alternative characterization procedure we have presented allowed us to measure the spectral features of the system with an accuracy comparable to standard laboratory equipment.

Smartphone-based imaging systems for medical applications: a critical review

“…On the other hand, many systems prototyped as “low-cost” systems on smartphones can likely be implemented at even lower fabrication cost with greater reproducibility using single board computers or embedded processors connected to peripheral camera sensors. 80 , 139 – 143 Therefore, the burden should be placed on SBI system developers to demonstrate the unique advantages of their systems through prospective clinical assessments. Achieving this will require working toward greater reproducibility and translation of SBI systems.…”

“…Rapid advances in computational imaging and optical design also hold great promise to be utilized in conjunction with existing smartphone optical sensors or to be added through additional dedicated sensors. 79 – 84…”

Evaluation of a melanoma screening framework based on depth resolved light scattering

“…With this setup, SFDS can measure the optical properties over a wide range of wavelengths in a broad spectrum. The SFDS system that we use in our experiments measures the optical properties ranging from 450 nm to 1050 nm in a single exposure [30][31][32][33]. SFDS may not be as precise or cover as large of an area as SFDI, but SFDI is usually limited in that it can only measure the optical properties at a select number of specific wavelengths determined by the LEDs used or filters in front of the camera.…”

“…The first case is a result of the SFDS/I system's optical design, specifically its chromatic aberrations and depth of focus which vary in relation to the ideal picture plane. A digital micromirror device which is tailored for wide field projection applications is normally used in the design process of these systems since many of these systems are constructed around pre-existing projection units [30,[54][55][56][57]. As a result, additional lenses are used to reduce the projection field of view but also reducing the working distance between the projection lens and the target to be measured which is beneficial towards small handheld systems in the future.…”

MEDUSA: An open-source and webcam based multispectral imaging system

“…Specific light band reflection results in “optical fingerprints” on an object image. While getting spectral information from RGB bands might be limited by a wide spectral band (full width at half maximum -FWHM- above 30 nm) and band superposition, cross RGB channel information from consumer grade cameras can perform like a standard grade spectroscopy system at analyzing medical samples [5] .…”