Single-shot Hyperspectral Fluorescent Imaging

2016 Fiscal Year Research-status Report

• Project/Area Number: 16K16095
• Research Institution: National Institute of Informatics
• Principal Investigator: 鄭 銀強 国立情報学研究所, コンテンツ科学研究系, 助教 (30756896)
• Project Period (FY): 2016-04-01 – 2018-03-31
• Keywords: Spectral Imaging / Fluorescence / Component Decomposition / Illumination Design

Outline of Annual Research Achievements

The core research objective of this project is to develop single-shot color imaging methods for fluorescent and reflective scenes. It has two major stages, including the usage of a single hyperspectral image (1st stage, fiscal year 2016-2017) and that of a multichannel image (2nd stage, fiscal year 2017-2018). In the fiscal year of 2016-2017, as the sole principle investigator, I have achieved the following research results:
(1). Developed a linear separation method of fluorescent and reflective components. By introducing the PCA based linear expression, the separation task boils down to a linear system, for which fast matrix operation tools exist. It is thus quite different from existing nonlinear separation methods.
(2). Mathematically designed the optimal illumination spectrum that suits best to stable and accurate decomposition. Thanks to the aforementioned linear separation, the optimal illumination can be designed by minimizing the condition number of the parametric coefficient matrix. I have designed a mathetical formulation for this condition number minimization problem.
(3). Found a low-cost commercial lamp that approximates the optimal illumination. Rather than manufacturing a new lamp, I have tried to investigate the spectra of dozens of commercial lamps, and found a HID lamp that properly approximates the optimal illumination.
(4). Applied the separation method and system to analyze coral samples in Okinawa, and verfied their robustness and applicability.

Current Status of Research Progress

1: Research has progressed more than it was originally planned.

Reason

The planned research topics (of developing a simple single-shot fluorescence/reflectance separation method, designing the optimal illumination, and finding a real and low-cost lamp) have been properly achieved, as mentioned in the achievements (1),(2),(3) above. The developed method and imaging system have been verified by applying them to analyze real coral samples in Okinawa. By collaborating with a biological researcher, we have successfully separated out the fluorescent components from several live and dead corals. Later, we examined the fluorescent protein structure and tried to link the structure with spectral distribution. As pre-research for the 2nd stage, I have also developed a three-channel imaging method for visibility enhancement of weak fluorescent substances.

Strategy for Future Research Activity

I plan first to summarize the current achievements and prepare a journal paper for submission. As for the planned research of multispectral fluorescent imaging, I am planning to investigate the following two topics:
(1). To examine the feasibility of separating fluorescent and reflective components in RGB trichromatic domain. If single-image based separation is intrinsically ambiguous, I plan to utilize the multiplexing effect of a DLP projector. In this case, realtime capture and separation might become possible for the first time.

(2). To better separate weak fluorescence for food freshness examination. Rotten food usually emits weak fluorescence, which can not be separated by existing methods. Our current fluorescence visibility enhancement method has potential in resolving this issue.