fluor 3

8/7/2019 fluor 3

1/6

Application of fluorescence lifetime imaging(FLIM) in latent finger mark detection

L.K. Seah *, P. Wang, V.M. Murukeshan, Z.X. Chao

School of Mechanical and Aerospace Engineering, Nanyang Technological University,

50 Nanyang Avenue, Singapore 639798, Singapore

Received 14 April 2005; received in revised form 1 August 2005; accepted 24 August 2005

Available online 22 September 2005

Abstract

Fluorescence lifetime imaging (FLIM) in frequency domain enables the mapping of the spatial distribution of fluorescence

lifetimes of a specimen. It has been extensively applied in biology. In this paper, a theoretical analysis for the fluorescence

lifetime determination of latent finger mark samples is described, which is followed by the feasibility study of using FLIM in

frequency domain for latent finger marks detection. Preliminary experiments are carried out with latent finger marks treated with

a fluorescent powder on two different substrates. The resulting fluorescence lifetime image of finger mark revealed a good

contrast, and was able to detect the latent finger marks clearly.

# 2005 Elsevier Ireland Ltd. All rights reserved.

Keywords: Fluorescence lifetime imaging; Frequency domain; Latent finger mark; Homodyne

1. Introduction

Finger marks are one of the most valuable types of

physical evidence and hence its detection plays a significant

role in criminal investigation and forensic science. There are

three types of finger marks that occur at a crime scene:

visible, plastic and latent. Latent finger mark is the most

difficult to be detected. The existing detection techniques for

latent finger marks have their own limitations. Most con-

ventional detection methods (physical and chemical) cannot

detect older marks [1,2]. Laser induced fluorescence detec-

tion of latent finger mark was initially explored in 1976 and

involves the detection of fluorescence intensity, color and

lifetime [9]. In general, conventional methods for detectionof latent finger marks take the first two properties into

consideration. However, the existing techniques based on

fluorescence filtering fail when there is strong fluorescence

emission from the background, and are ineffective when the

emission wavelengths from the finger mark and that of

background fall in the close wavelength range. In this

context, FLIM method would be beneficial.

Fluorescence lifetime is the average decay time of the

fluorescence emitted by a molecule after excitation with a

short laser pulse. Fluorescence lifetime imaging (FLIM)

has received considerable attention and has been widely

used in biophysics and medical diagnosis [36]. FLIM

could obtain the unique and quantitative information

available from dynamic fluorescence measurements, and

possibly distinguish several fluorescence species with

dissimilar lifetimes even though they may have overlap-

ping spectra.

This paper deals with determination of average fluores-cence lifetimes of latent finger mark sample on a pixel-by-

pixel basis using homodyne detection method. A number of

images of the sample are collected at different phase delays

relative to the excitation light. A subsequent fit of fluores-

cence intensity on a pixel-by-pixel basis yields the fluores-

cence lifetime distribution image of the latent finger mark,

which is independent of the fluorescence intensity.

www.elsevier.com/locate/forsciintForensic Science International 160 (2006) 109114

* Corresponding author. Tel.: +65 6790 4824; fax: +65 6795 4632.

E-mail address: [email protected] (L.K. Seah).

0379-0738/$ see front matter # 2005 Elsevier Ireland Ltd. All rights reserved.

doi:10.1016/j.forsciint.2005.08.018

8/7/2019 fluor 3

2/6

2. Principle and theory of fluorescence lifetime

measurement

2.1. Fluorescence lifetime measurement in frequency

domain

The measurement of fluorescence lifetime is based on the

cross-correlation method introduced by Spencer and Weber

[7] (Fig. 1). The lifetime resolved fluorescence experiment is

carried out by modulating the excitation light sinusoidally at

a high frequency (typically on order of 10100 MHz) and

determining the demodulation and phase of high frequency

modulated fluorescence emission.

The fluorescence lifetime can be calculated by measuring

the phase shift and demodulation of the fluorescence emis-

sion relative to the phase and demodulation depth of the

excitation light.

If the excitation light intensity is modulated as a sinu-

soidal function of time Ev(t) with a frequency f, the fluor-escence emission intensity is a sinusoidal function with the

same frequency [8]:

Evt E0;v1 ME;vsinv t fE;v (1)

Fvt F0;v1 MF;vsinv t fF;v: (2)

where E0,v is the time-independent offset, ME,v the relative

modulation of the excitation light, v the radial frequency

(v = 2pf)andfE,v is the arbitrary phase lag of the excitation.

The symbols F0,v, M0,v, MF,v and fF,v of Eq. (2) are similar

but refer to the fluorescence emission. The total phase lag of

the fluorescence emission relative to the excitation light Df

and modulation Mcan be expressed by

Df fF;v fE;v (3)

MMF;v

ME;v(4)

For a single lifetime, the lifetime can be determined from

both Df and M:

tf 1

vtanDf (5)

tM1

v

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi1

M2 1

r(6)

where tf and tMare the lifetimes calculated from Df and M,

respectively. If they aredifferent, it can be concluded that the

system is heterogeneous. In this paper, the average lifetime

was calculated by Eq. (5).

The appropriate modulation frequencies for measuring

fluorescence lifetimes, which fall in the range of 110 ns, are

20100 MHz. It is impossible to directly measure the phase

and demodulation of the time-varying signals with such high

frequency. Therefore, the high frequency signal is trans-

formed to either DC (homodyne measurement) or very low

frequencies (heterodyne measurement), hence the phase and

demodulation can easily be measured. In this paper, homo-

dyne method was used as it is relatively simple compared to

heterodyne method.

2.2. Homodyne detection

The principle of the homodyne measurement can be

introduced easily for the simplest case of a perfectly sinu-soidal excitation and a single fluorescence lifetime. For

homodyne detection, the source and detector frequencies

are the same. The intensity is proportional to the sine

function of the phase difference between the detector and

the emission. To acquire the entire phase and modulation

information, it is necessary to change the phase of the

detecting device relative to that of the excitation light. In

most of the case, the gain of the detection is modulated

according to

Gvt G0;v1 MG;vsinv t fG;v: (7)

where G0,v is the time-independent offset, MG,v the relative

modulation of the gain and fG,v is the arbitrary phase lag ofthe gain.

Then the detected intensity is given by

F0t F0;01 MF;0sinf0G;v fF;v; (8)

where F0;0 F0;vG0;v; MF;0 12MF;vMG;v andf

0G;v

fG;v p2:

To obtain Df and M, a reference measurement must be

made by measuring a reference compound with known

fluorescence lifetime (tR), the detected signal has the similar

expression

R0t R0;01 MR;0sinf0G;v fR;v (9)

where R0;0 R0;vG0;v andMR;0 12 MR;vMG;v:Then Df and Mare defined as

Df fF;v fs;v fF;v fR;v tan1vtR (10)

MMF;v

MS;v

MF;v

MR;v

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi1 vtR

2q (11)

The subscript Sdenotes scattered light whose effective

lifetime is zero.

L.K. Seah et al. / Forensic Science International 160 (2006) 109114110

Fig. 1. Phase shift Df and demodulation Mof the fluorescence

signal with respect to the excitation light. The modulation

M= (ACem/DCem)/(ACex/DCex).

8/7/2019 fluor 3

3/6

2.3. Data analysis

The data are taken as a series of subimages collected at

equidistant phase steps. For Eqs. (10) and (11), the key is to

get the value ofMF,v and fF,v. For a convenient analysis,

Eq. (8) is redefined as:

yt;i a11 a2sinxi a3 (12)

where a1 = F0,0, a2 = MF,0, a3 = fF,v and xi f0G;v;i,

where f0G;v;i is the different phase value of the gain. The

goal of curve fitting is to find the value ofa1, a2 and a3 that

will make the calculated results yt,i closest to the detected

signals yd,i [8]. Eq. (12) is a non-linear relationship, but a

linear expression could be obtained through transformation:

yt;i c1 sin xi c2 cos xi c3 (13)

where c1 a1a2 cos a3; c2 a1a2 sin a3 and c3 a1:In order to do the linear curve fit, Eq. (13) was regarded as

multi-variable function. The images include the informationofyd,i and xi. After the process of multi-variable curve fit, the

values of coefficients c1, c2 and c3, are obtained. The values

ofa1, a2 and a3, are calculated by

a1 c3

a2

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffic21 c

22

pc3

a3 tanc2

c1

b1 > 0

a3 tanc2

c1

p b1 < 0

8>>>:

(14)

Through the same process, a2,R and a3,R are calculated.

The Mand Df are obtained by simple calculation:

Ma2;F

a2;R

1ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi1 vtR

2q (15)

Df a3;F a3;R tan1vtR; (16)

where Fand R refer to the sample and the reference signal,

respectively.

3. Experimental setup and procedure

The schematic diagram of the experimental set up is

shown in Fig. 2. A CW Argon laser (Coherent Innova 90 C)

is modulated using an electro optic modulator (Conoptics

EOM, Model 350-210) capable of operating in the range of

0100 MHz. A power amplifier is used as the power source

to the modulator, which is modulated by a sinusoidal signal

output from function generator 1 (Agilent, Model 33250A).

The modulated laser light is coupled into a single mode

optical fiber (OZ optics) to illuminate the sample treated

with fluorescence powders. An intensified charge coupled

device camera (ICCD, Lavison, Pico Star HR) collects the

resulting fluorescence emission from the sample. Function

generator 2 (Agilent, Model 33250A) is used as the signalsource to modulate the gain of ICCD. The two function

generators aresynchronized. A long pass optical filter is used

to cut off the excitation-scattered light from the sample. The

ICCD is connected with a camera interface board and

controlled by a computer.

By changing the phase of the function generator 2, the

phase shift of the detecting device relative to the exciting

light is changed. In order to have a good fit, 10 images are

obtained in one period equal spaced out of 368 Then the

different phase step of the gain is defined as:

X 0;p=5; 2p=5; 3p=5; 4p=5;p; 6p=5; 7p=5; 8p=5; 9p=5

After 10 sample images and 10 reference images are

collected at equidistant phase steps, they are used to calcu-

late the lifetime by FLIM program, which is implemented in

MATLAB. The program is outlined schematically in Fig. 3.

First a2,R and a3,R are obtained by calculating the reference

data. After importing the source data, for every pixel whose

L.K. Seah et al. / Forensic Science International 160 (2006) 109114 111

Fig. 2. Schematic diagram of the experiment setup.

8/7/2019 fluor 3

4/6

intensity is within the threshold, the intensity values of

equidistant phase steps are fitted to a sine wave to yield

the demodulation factor a2,Fand the phase shift factor a3,F.

In order to process the curve fitting procedure, Eq. (13) is

transformed to

Y c1

sinX c2

cosX c3

(17)

The function in MATLAB toolbox is used to obtain the

coefficients. Eq. (14) is used to obtain the value ofa1,F, a2,F

and a3,F. Then MandDf arecalculated to decide the lifetime

through Eqs. (15) and (16). After every pixel is processed,

the maximum and minimum values are set as the value range

to output the lifetime image.

In order to simulate a condition of having very close

emission wavelengths for finger mark and background

fluorescence, fresh finger marks from a single person are

deposited on fluorescing color paper and treated with blitz-

green powder (Lightning Powder Company) that is applied


Fig. 3. Block diagram of FLIM program.

Fig. 4. The intensity image (a) and fluorescence lifetime image (b) of the blitz-green-treated finger mark samples on green fluorescence paper

substrate.

8/7/2019 fluor 3

5/6

by means of a magneticbrush.The emissionwavelength of the

fluorescing color paper is 516 nm and its lifetime is 3.1 ns,

while those of blitz-green powder are 523 nm and 8.96 nm.

Experiment was also carried out on the finger mark

treated with blitz-green fluorescent powder on smooth

calendar paper. The background has two major fluores-

cence components with lifetime value 13.42 ns and

2.56 ns. Lifetime value of finger mark sample treated

with blitz-green powder is 8.96 ns, which is shorter than

that of the major fluorescence component (13.42 ns) from

the background

4. Result and discussion

Selecting a 700 700 pixel region, the intensity imageof the finger mark on green fluorescence paper treated

with blitz-green powder is shown in Fig. 4a. The contrast

of the latent finger mark image is found to be poor and

hence could not identify the ridge details. This is due to

that the emission wavelength of green fluorescence paper

(516 nm) is merged with that of the latent finger mark

treated with blitz-green fluorescent powder (523 nm).

Fig. 4b shows the corresponding average lifetime image

of latent finger mark. The fluorescence powder has longer

lifetime than the background fluorescence of the green

paper. It can be seen that lifetime image effectivelysuppress the background fluorescence so as to give a clear

ridge details of the finger mark

Fig. 5 shows the intensity image and fluorescence life-

time image of the finger mark treated with blitz-green

powder on postcard. Due to the strong multiple fluorescence

emissions from the background, the latent finger mark

cannot be identified under normal imaging. As mentioned

in the previous case, in the fluorescence lifetime image the

background can be subdued successfully despite the back-

ground fluorescence is having a higher intensity emission.

For the case that the background has longer lifetime, a

negative image will be obtained compared to Figs. 4b and 5b.

The finger mark can also be detected in that case.

5. Conclusion

FLIM has been widely used in biology and biomedical

fields. In this paper, lifetime imaging of latent finger marks

on two different substrates were carried out. An approach

that involved the calculation of total phase lag and demo-

dulation factor is used to determine the lifetimes pixel by

pixel. The results show that the FLIM can detect the latent

finger mark whose emission wavelength falls in the close

wavelength range with that of the background. The ridge

details of latent finger mark can be identified clearly. The use

of lifetime imaging yields results that are independent of the

fluorescence intensity. FLIM technique is a general approach

for latent finger mark detection and is only limited by the

lifetime difference between the substrate and the finger

mark. Hence, as long as there is a lifetime difference that

can be detected, this technique can be applied.

Acknowledgements

The authors gratefully acknowledge the support and fund-ing from The Enterprise Challenge (TEC), Singapore and

Academic Research Fund (AcRF), Nanyang Technological

University. The authors express their sincere gratitude to

Singapore Police Force for their collaboration in this project.

References

[1] E.R. Menzel, Fingerprint Detection with Lasers, second ed.,

Marcel Dekker, New York, 1999.

L.K. Seah et al. / Forensic Science International 160 (2006) 109114 113

Fig. 5. The intensity image (a) and fluorescence lifetime image (b) of the blitz-green-treated finger mark samples on postcard substrate.

8/7/2019 fluor 3

6/6

[2] H.C. Lee, R.E. Gaensslen, Advances in Fingerprint Technology,

Elsevier, New York, 1991.

[3] R. Sanders, H.C. Gererritsen, Fluorescence lifetime imaging of

free calcium in single cells, Bioimaging 2 (1994) 131

138.

[4] D. Elson, S. Webb, J. Siegel, K. Suhling, D. Davis, J. Lever, D.Phillips, A. Wallace, P. French, Biomedical applications of

fluorescence lifetime imaging, Opt. Photon. News 11 (2002)

2732.

[5] K. Dowling, M.J. Dayel, Hyde SCW, French PMW, M.J. Lever,

J.D. Hares, High resolution time-domain fluorescence lifetime

imaging for biomedical applications, J. Mod. Opt. 46 (2) (1999)

199209.

[6] M. Elangovan, R.N. Day, A. Periasamy, Nanosecond fluores-

cence resonance energy transfer-fluorescence lifetime imaging

microscopy to localize the protein interactions in a single living

cell, J. Microsc. 205 (2002) 314.

[7] R.D. Spencer, G. Weber, Measurements of subnanosecond

fluorescence lifetime with a cross-correlation phase fluorometer,Ann. N. Y. Acad. Sci. 158 (1969) 361376.

[8] Gadella TWJ, R.M. Clegg, T.M. Jovin, Fluorescence lifetime

imaging microscopy pixel-by-pixel analysis of phase-modula-

tion data, Bioimaging 2 (1994) 135159.

[9] L.K. Seah, U.S. Dinish, S.K. Ong, Z.X. Chao, V.M. Muruke-

shan, Time-resolved imaging of latent fingermarks with nano-

second resolution, Opt. Laser Tech. 36 (5) (2004) 371376.


fluor 3

Documents