Algorithm validation using multicolor phantoms

Daniel V. Samarov,Ji Youn Lee,Jeeseong Hwang,Maritoni Litorja,Matthew L. Clarke,David W. Allen

doi:10.1364/boe.3.001300

Abstract

We present a framework for hyperspectral image (HSI) analysis validation, specifically abundance fraction estimation based on HSI measurements of water soluble dye mixtures printed on microarray chips. In our work we focus on the performance of two algorithms, the Least Absolute Shrinkage and Selection Operator (LASSO) and the Spatial LASSO (SPLASSO). The LASSO is a well known statistical method for simultaneously performing model estimation and variable selection. In the context of estimating abundance fractions in a HSI scene, the “sparse” representations provided by the LASSO are appropriate as not every pixel will be expected to contain every endmember. The SPLASSO is a novel approach we introduce here for HSI analysis which takes the framework of the LASSO algorithm a step further and incorporates the rich spatial information which is available in HSI to further improve the estimates of abundance. In our work here we introduce the dye mixture platform as a new benchmark data set for hyperspectral biomedical image processing and show our algorithm’s improvement over the standard LASSO.

Highlights

Hyperspectral imaging (HSI) is a technology commonly used in remote sensing but has recently found increased use in biomedicine, from investigations at the cellular [1] to the tissue level[2, 3] and often captured in reflectance mode
In the work presented here we focus on endmember abundance fraction estimation
The vertical white lines delineate the different dye mixtures shown at the bottom and the color bar to the right of each figure shows the range of the estimated abundance fractions

Summary

Introduction

Hyperspectral imaging (HSI) is a technology commonly used in remote sensing but has recently found increased use in biomedicine, from investigations at the cellular [1] to the tissue level[2, 3] and often captured in reflectance mode. Measurement accuracy depends on radiometric proficiency and the robustness of the algorithm used to extract the desired quantity Radiometric challenges are both instrument-related (e.g. drift, noise, sensor inhomogeneities) and scene-related (e.g. glare, topography, light field non-uniformity). The algorithm on the other hand, needs to accurately extract the analyte quantity in the presence of signal contaminants such as scattering and signal interferers arising from undesired absorption and spatial and spectral correlations It requires validation against known preparations, closely resembling in properties to the real samples. Tissue phantoms are prepared for this purpose[4] These ground-truth artifacts tend to have short shelf life and are often prepared with other functional considerations besides optical property.

Methods

Results

Conclusion