Note: Descriptions are shown in the official language in which they were submitted.
CA 02796489 2012-11-22
Man-Portable Device for Detecting Hazardous Material
Field of the Invention
This invention relates to the field of hazardous material detection, and in
particular
to a portable for detecting the presence of hazardous material, such as
Chemical,
Biological, Radioactive, Nuclear, Explosive (CNBRE) material in the field.
Background of the Invention
Current practice in the field for the detection of very low concentrations of
threatening CBRNE surface contaminants is to sweep the target surface with a
special cloth known to be free of all CBRNE traces. Once the targeted surface
has
been swept, the cloth is placed in a sealed container for subsequent (bio-)
chemistry
analysis. Only one surface is swept with a given cloth to avoid dilution of
the sample
material. The (bio-) chemistry procedures for a given sample may take several
minutes up to several days depending on the complexity of the analysis
process.
These (bio-) chemistry procedures have the advantage of providing various
depth of
information (usually correlated with the time and complexity of the executed
procedure) on the sampled material that can be tailored to the objectives of
the
CBRNE contamination survey (presumptive detection up to forensic
investigation).
There are two major limitations associated with the physical sampling of a
given
surface with a dedicated cloth and a subsequent timely demanding (bio-)
chemistry
procedure in assessing the presence of very low concentration of CBRNE
contaminants. First, investigating a CBRNE crisis event at a given location
involves
blindly sampling numerous surfaces for the presence of contaminants. Each
sample
1
CA 02796489 2012-11-22
requires a single cloth and a subsequent (bio-) chemistry procedure. In some
cases
where the nature of the CBRNE event is unknown, multiple (bio-) chemistry
procedures may be performed from a single sample. This blind investigation
requires important man-powered efforts, a sizeable equipment infrastructure
and a
non negligible quantity of consumables while usually only a very small
fraction of
the processed samples will disclose the presence of CBRNE contaminants.
Second, a
non-negligible delay is unavoidable between the sampling procedure and the
outcomes of the (bio-) chemistry analyses. This delay may result in a delay in
taking
important decisions concerning the rapid initiation of efficient protection
measures,
the adequate delimitation of the prohibited areas and the quick selection of
adequate medical countermeasures.
US patent publication no. 2012/0038908 discloses a system for detecting CNBRE
hazards using fluorescence and spectroscopic imaging techniques wherein data
sets
generating during interrogation are compared to a reference database. However,
a
problem is that each specimen may contain several tens of thousands of samples
acquired within a few seconds, and producing a rapid response can be a serious
challenge.
US patent no. 6,272,376 discloses the use of time-resolved, laser-induced
fluorescence spectroscopy in the identification of tissue.
US patent no. 7,821,633 discloses an apparatus for performing Raman
spectroscopy
with time domain spectral analysis.
2
CA 02796489 2012-11-22
US patent no. 8,190,242 discloses a laser synthesizer for multi-dimensional
spectroscopy.
Summary of the Invention
Embodiments of the invention provide a tool for the First Responders to
identify
rapidly the most promising surfaces that may present CBRNE contaminants, to
reduce greatly the number of surfaces where physical sampling is performed and
the subsequent (bio-) chemistry procedures, and to provide within seconds a
level
of classification of the detected CBRNE contaminants allowing the immediate
selection of adequate protective measures.
The invention combines time-resolved spectrofluorimetry and multi-variate
statistical analysis in a single field instrument well adapted to First
Responders
Operating Concepts when having to investigate for the presence of very low
concentration CBRNE surface contaminants at a reported crisis site.
According to the present invention there is provided a man portable device for
detecting the presence of hazardous material, comprising: a pulsed or time-
modulated light source; an objective scannable across the surface of a sample
for
projecting light from the light source onto a succession of spots on the
surface of the
sample to induce fluorescence; a spectrometer for performing a spectral
analysis of
the induced fluorescence to create a first dataset defining a first vector as
a function
of wavelength; a time domain detector for measuring the time decay of the
induced
fluorescence to create a second dataset defining a second vector as a function
of
time; and a computer configured to identify hazardous material by performing
3
CA 02796489 2012-11-22
independent multivariate analysis on the first and second vectors as the
objective is
scanned across the sample surface based on fluorescent signal models for
hazardous
materials in the spectral and time domains.
The fluorescent signal models may be stored in memory and based on theoretical
or
experimental data or a combination of the two.
The combination of spectrally-resolved UV light induced fluorescence and time-
resolved fluorescence decay resulting from probing statistically independent
tiny
areas part of the inspected surface, is used to detect, classify and quantify
very small
concentration of fluorescing CBRNE surface contaminants.
In order to minimize the clutter generated by background surfaces and to
address
the sparse distribution of micron size contaminants on these surfaces, the
instrument sweeps a targeted surface using dedicated spacers and fires a
series of
light pulses on areas of this surface ranging from a few to a few tens of
microns in
size. The manual sweeping speed, size of the irradiated areas and frequency of
the
light pulses are adjusted to probe a different area with each pulses. Through
a single
sweep, an ensemble of tens, if not hundreds of thousands of spots or areas is
probed.
The spectrally and time resolved light induced fluorescence (LIF) associated
with
each areas are analyzed with a multi-variate algorithm. Based on a library
containing the LIF spectral and temporal characteristics of multiple products
of
interest, the detection/classification of the compounds among the end-members
of
this library as well as the quantification of their concentrations on the
probed
surface are derived from the multi-variate analysis of this ensemble. This
4
detection/classification/quantification results are communicated within
seconds to the
First Responders through a simple small electronic display.
In one aspect, there is provided a man portable device for detecting the
presence of
hazardous material, comprising:
a pulsed or time-modulated light source;
an objective scannable across a surface of a sample for projecting light from
the light
source onto a succession of spots on the surface of the sample to induce
fluorescence;
a spectrometer for performing a spectral analysis of the induced fluorescence
to create a
first dataset defining a first vector as a function of wavelength;
a time domain detector for measuring the time decay of the induced
fluorescence to
create a second dataset defining a second vector as a function of time; and
a computer configured to identify hazardous material by performing independent
multivariate analysis on the first and second vectors as the objective is
scanned across the
sample surface based on fluorescent signal models for hazardous materials in
the spectral
and time domains.
In one aspect, there is provided a computer-implemented method of detecting
the
presence of hazardous material, comprising:
scanning a pulsed or time-modulated light source, using an objective, across a
surface of
a sample onto a succession of spots on the surface of the sample to induce
fluorescence;
performing a spectral analysis of the induced fluorescence to create a first
dataset
defining a first vector as a function of wavelength;
measuring the time decay of the induced fluorescence to create a second
dataset defining
a second vector as a function of time; and
CA 2796489 2018-10-22
identifying hazardous material by performing independent multivariate analysis
on the
first and second vectors as the objective is scanned across the sample surface
based on
fluorescent signal models for hazardous materials in the spectral and time
domains.
According to another aspect of the present invention, there is provided a
computer-
implemented method of detecting the presence of hazardous material,
comprising:
scanning a pulsed or time-modulated light source, using an objective, across a
surface of a sample onto a succession of spots on the surface of the sample to
induce
fluorescence;
performing a spectral analysis of the induced fluorescence, using a
spectrometer,
to create a first dataset defining a first vector as a function of wavelength;
measuring the time decay of the induced fluorescence, using a time domain
detector, to create a second dataset defining a second vector as a function of
time; and
identifying hazardous material by performing independent multivariate analysis
on the first and second vectors as the objective is scanned across the sample
surface
based on fluorescent signal models for hazardous materials in the spectral and
time
domains.
Brief Description of the Drawings
The invention will now be described in more detail, by way of example only,
with
reference to the accompanying drawings, in which: -
Figure 1 is a block diagram of a man-portable device in accordance with an
embodiment
of the invention; and
Figure 2 is an example of an exponential transient decay associated with
induced
fluorescence.
5a
CA 2796489 2019-04-01
Detailed Description of the Invention
As shown in Figure 1, the instrument is composed of two main hardware
components: the
optical probe 10 and the electronic enclosure 12, the two being linked with
optical fibers
and low power electric wires.
The optical probe 10 includes the optical multiplexing assembly, which may
include cube
beamsplitters, semi-transparent mirrors, dichroic mirrors, dichroic
beamsplitters, volume
Bragg grating and holographic beam samplers as well as intermediate lens and
optical
filters as shown in Figure 1. The optical multiplexing assembly combines and
separates
the different optical signals going through the objective. It routes the
excitation light
pulses toward the surface under investigation and the returned fluorescence or
reflected
imaging light to the spectrometer 24 (via optical fiber 36), the transient
detector 48, and
the camera 40:
5b
CA 2796489 2019-04-01
CA 02796489 2012-11-22
A simple small electronic display 14 is located on the optical probe 10 and
the
combined component is ergonomically designed as a pen-like probe to be
manipulated as a large pen. A dedicated spacer 16, based on ball bearing or an
equivalent technology facilitating the sweeping process, is in contact with
the
probed surface 18 and keeps approximately constant the distance between the
pen's optical port and the probed surface during the sweep. Special care is
applied
to the design of this spacer component 16 to eliminate all possible subsequent
contaminations between sweeps or after the investigation process.
The electronic enclosure contains a battery 20, an excitation light source 22
(if not
integrated to the optical probe), the imaging spectrometer 24, the high speed
transient recorder electronics 26 as well as the computing components 28
controlling the instruments, recording the collected information, executing
the
multi-variate program analysis over the ensembles of collected data,
presenting the
results on the small electronic display 14 and controlling specialized data
communication through electronic data ports 50. The computer 28 is connected
to a
memory 52 storing a library of spectral and time domain signatures based on
models for hazardous materials in the spectral and time domains. The
electronic
enclosure 12 is sufficiently small and low weight to be located in a man-
portable
packsack and can be operated for a few hours without recharging the battery.
The excitation light from source 22 is passed over optical fiber 30 (or
directly from a
small light source 53 powered by power supply 22) to dichroic filter 32 in the
optical probe 10, where it is redirected to the objective 19 maintained at a
fixed
distance above the surface 18 by the spacer 16. Light returned from the
objective 19
6
CA 02796489 2012-11-22
is split by a first beam splitter 34, where a portion is directed to imaging
spectrometer 24 over optical fiber 36 and another portion is directed to
second
beam splitter 38, which in turn divides the beam into a portion directed to
camera
40 electrically connected via 42 to computer 28 and a portion coupled over the
transient detector 48 electrically connected to the transient recorder 26.
The objective 19 is an optical assembly used to image the surface to be
analyzed,
focus the fluorescence inducing source 53 and collect the induced fluorescence
signal. Since the biological contaminant could be single biological specie,
the
objective should be able to resolve micron size particles.
The objective 19 should meet a number of requirements, such as having a good
transmission for the excitation light source and for the induced fluorescence
signal.
It must also be constructed so that its optical materials produce very limited
or near
zero auto-fluorescence. It must be relatively achromatized, meaning that its
focal
length has little dependency on the wavelengths.
Objectives available commercially can be based on refractive or reflective
optics.
There are also hybrid objectives based on both refractive and reflective
optics.
Another type uses diffraction, as holographic lenses, to achieve more or less
the
same imaging function.
It is desirable to have an objective having a medium focal length, preferably
between 5 and 12 mm, but objectives with 2 to 40 mm focus distances can also
achieve satisfactory results.
7
CA 02796489 2012-11-22
Another characteristic of importance for practical reasons is the need for a
long
working distance. This last characteristic minimizes the probability of
contaminating the probe (beside the dedicated spacer) by a contact between the
objective and the probed surface. A long working distance is also useful to
avoid
contamination if the system incorporates an automatic focus function that
varies the
distance between the objective and the probed surface. The numerical aperture
of
the objective should be chosen as high as possible to maximize the probe's
light
collecting ability and imaging resolution.
The materials of the objective's lenses should be chosen to minimize losses at
the
excitation wavelength and in the spectral band of the anticipated induced
fluorescence from the diverse surfaces and compounds to be encountered.
Finally,
for an optimum optical design, the chosen objective should be of the infinity-
corrected type. This type of objective outputs parallel light rays originating
from
points at the probed surface once focused, or equally said, images points from
the
probed surface at infinity. This characteristic greatly facilitates the
insertion of
subsequent optical components as beam splitters, lenses or mirrors along the
optical path without affecting significantly the achievable resolution. With
such an
objective, the light collected from the probed surface is put in a bundle of
rays that
will pass through subsequent lenses or mirrors and be focused at their exact
focal
lengths. This also defines the imaging system magnifications as the ratio of
the focus
distances between the subsequent lenses/mirrors and the objective. An example
of
an objective that we have used and meets the different requirements is the
Nikon
10X CFI LV plan EPI lens.
8
CA 02796489 2012-11-22
In order to be able to perform time-resolved spectrofluorimetry on
contaminated
surface, the excitation light source originating from either a source 22
situated in
the electronic enclosure 12 or 53 in the optical probe 10 should be a pulsed
or time
modulated light source. In the described embodiment, a pulse laser is chosen
but
other pulsed optical sources such a light emitting diode (LED) could also be
used.
Because the anticipated induced fluorescence should have duration of a few
nanoseconds or more, the excitation light pulse duration should be of the
order of 1
nanosecond or less to produce quasi-instantaneous induced fluorescence
response.
To construct an ensemble composed of several tens if not hundreds of thousands
of
induced fluorescence spectral and time decay samples with a single sweep
lasting a
few seconds, the excitation light source should possess a pulse repetition
frequency
(PRF) of 10-100 kHz or higher.
The produced light may be coupled to an optical fiber with fiber tips located
at
position 53, or, if the source is sufficiently compact, having the source
directly
inserted at that location. The excitation light rays emitted from this
position are
collected by a lens and directed through the objective 19 and focused on the
surface
18 to be analyzed.
The dichroic beamsplitter 32 is used to reflect and direct the excitation
light through
the objective while being transparent to the induced fluorescence.
The choice of the excitation light source spectrum is important as it defines
the
specific chromophores present in the researched contamination that will be
excited
and produce the fluorescent signal upon which the time-resolved
9
CA 02796489 2012-11-22
spectrofluorimetry will be performed. Embodiments of the invention can also
accommodate a multi-wavelength excitation light source that fires either
simultaneously or sequentially light pulses containing a multiple number of
wavelengths to extend the characterization of the observed surface and its
contaminations by exciting a larger number of chromophores.
For a single wavelength excitation, a UV wavelength light source at around 355
nm
could be used. It is well known that an excitation at this wavelength is
capable of
exciting NADH, a fluorescent compound present in living bacteria. The
fluorescence
given by the sampled compound on the surface is then spectrally and
dynamically
(temporally) measured giving both signatures, which can be classified, with
each
pulse fired by multivariate analysis. The fluorescence of NADH present in a
bacterium has characteristic spectrum and time decay that occur mainly in the
400-
540 nm bands. Therefore, such fluorescent response in these bands having the
corresponding characteristics would indicate the presence of this class of
biological
contaminants on the probed surface. As an example, to further assess the class
of
that biological contaminant, a second light excitation at 266 nm may be fired.
This
wavelength is known to be able to induce a fluorescence signal from
tryptophan,
another substance present in bacteria. The fluorescence of tryptophan occurs
in the
310-420 nm bands. This additional fluorescence response (spectrometric and
dynamic) may further improve the probability of a valid classification of the
contaminants. Different types of bacteria have different relative contents of
NADH
and tryptophan producing different signatures, which can be, used to further
detail
the classification results. Furthermore, there are other components in
bacteria able
CA 02796489 2012-11-22
to give fluorescent response to a UV excitation like tyrosine, phenylaniline
and
pyridine to name a few.
The overall fluorescence responses (spectral and dynamic) of a class of
bacteria
containing these multiple compounds with specific relative contents that
excite at
multiple wavelengths constitute the basis of a robust classification scheme.
To collect the various induced fluorescence, the collecting optics should
possess a
fairly good optical transmission between 300 and 700 nm. Additional dichroic
filters
may be added to further eliminate excitation light that may parasitize the
collected
fluorescence. Also, neutral filters may also be inserted along the excitation
and
collection optical paths to avoid contaminant bleaching or saturation of
detection
electronics.
Since biological materials and many other fluorescing substances of interest
present
good efficiencies (several percents) to produce fluorescence for a given
excitation
energy and that the areas of the probed surface irradiated by the UV light
pulses will
be fairly small (few to 50 tm in diameter) to minimize the fluorescence
contribution
of the background surface in comparison of micron size contaminants, the
overall
energy per light pulse will be fairly modest (a few tens ofp.J or less).
The imaging subsystem in the optical probe 10 provides the general
functionality
allowing the user to inspect a selection of fluorescence images, each taken
with a
single pulse during the sweep. Such a system is fairly simple to implement and
add
basic functionalities. One of these functionalities is to allow the user to
recall some
of the images captured by the camera 40, tagged with the multivariate analysis
of
11
CA 02796489 2012-11-22
the corresponding light pulses, to inspect the shape/size of the fluorescent
contaminants part of a same multivariate region of interest. It may also be
used to
visually inspect the analyzed surface as a classical fielded/portable
microscope. The
imaging subsystem is implemented by sampling with a partial mirror or a cube
beamsplitter 34, 38 a part of the ray bundle coming from the surface and using
a
lens to image this light bundle on the camera 40 having a multi-pixel sensor,
which
may be a CCD, ICCD or CMOS device.
This camera 40 is connected to an electronic system (shown in Fig. 1 as part
of the
computer) that gathers the signals from the camera and transforms it so that
it
could be put to a memory module for subsequent inspection and record keeping.
The camera 40 can be used in two modes. First, it can image the induced
fluorescence from the probed surface triggered by a single excitation light
pulse at a
fairly high repetition rate but most probably at a fraction of the one
associated with
excitation light pulse. The aperture duration of the camera must be
sufficiently
shorter than the time period between two light pulses to avoid contamination
between two probed areas. The second imaging mode uses rapid white lighting
pulses integrated to the optical probe to produce classic microscopy still
images of
the analyzed surface. These images may be recorded and/or directly send to the
electronic display for live visual inspection of still images or movies of
that surface.
The use of a white short-pulsed lighting eliminates imaging blurs that may
result
from movement of the optical probe over the inspected surface. The rapid white
light pulses lighting the imaged area of the probed surface could be
implemented in
12
CA 02796489 2012-11-22
many ways, for example, with an electronic flash or a pulsed LED or LED arrays
surrounding the objective.
For the two imaging modes (sampling imaged LIF or classical visible images),
keeping the probed surface 18 within the field depth of the objective is
important.
This is particularly important for the analysis of structured surfaces having
a certain
level of roughness of having some curvature under a large magnification. This
can be
achieved with an autofocus function where the quality of the images produced
is
used as feedback information driving mechanically the vertical position of the
objective. There are nowadays many optical systems that use a camera sensor
and
an electronic circuit to give the require commands keeping the objective lens
at the
right focus distance from an analyzed surface.
The spectrometer subassembly constitutes an important part of the system as it
separates the wavelengths of the fluorescence induced from the probed surface.
This assembly is designed with a spectral resolution between 3 and 5 nm to
accommodate the usual wide fluorescence spectra (several tens of nanometers)
of
chromophores. It is composed of a lens focusing the collected induced
fluorescence
in the optical fiber tip 46. This optical fiber guides this collected
fluorescence to a
coupling lens of the imaging spectrometer 24. One end of that fiber is
positioned at
the focus of the lens of the optical probe, determining the induced
fluorescent
collecting field-of-view of the spectrometric subassembly, while the other end
co-
located with a lens adapting the f-number with the entrance imaging optics of
the
spectrometer.
13
CA 02796489 2012-11-22
The spectrometric subassembly magnification should be designed to have the
induced fluorescence collecting field-of-view comparable or slightly larger
than the
UV excitation light spot produced at the focus plane of the objective and the
two
must be co-centered. The size of the optical fiber core, the spectrometer
grating as
well as the spectrometer optical design are selected to achieve the targeted
spectral
resolution while keeping a fairly good throughput. A multi-pixel sensor is
positioned
at the exit-imaging plane of the spectrometer to detect and record separately
the
spectrum produced by each UV excitation light pulses. This multi-pixel sensor
must
readout and records each fluorescence spectrum at the pulse repetition
frequency
(PRF) of the excitation source. This multi-pixel sensor may be a photodiode
linear
array, an avalanche photodiode (APD) linear array or a photomultiplier tube
(PMT)
linear array. If the PRF is sufficiently low, CCD, ICCD, EMCCD or CMOS 2D-
detectors
arrays may be used instead.
The time domain detection subassembly is another important component of the
system as it gives a measurement of the time decay of the induced fluorescence
from
the different chromophores that have been excited by the UV excitation light
source.
It provides valuable dynamic signatures processed in parallel with the
spectral
signatures obtained simultaneously which, once combined, constitutes the
individual sample of the ensemble collected during the sweep and analyzed by
the
multivariate technique. The time domain detection subassembly is composed of
the
objective 19, the partially reflective beam splitters 34 and 38, and a lens
focusing
the sample fluorescence on a fast optical detector 48 as a photodiode, an APD
or a
PMT.
14
CA 02796489 2012-11-22
An optical fiber end may also be placed at the position 48 and have the fast
optical
detector located at the other end of the fiber within the transient recorder
26 of the
electronic enclosure 12.
As for the spectrometric subassembly, the time domain detection subassembly
must
be designed to have the induced fluorescence collecting field-of-view larger
than the
UV excitation light spot produced at the focal plane of the objective and the
two
must be co-centered. The anticipated time decay of biological contaminants is
of the
order of a few nanoseconds while other materials may show time decays much
longer (microseconds and up to seconds for certain minerals). This implies
that the
transient recording of the fluorescence decay should have a time resolution of
a
nanosecond or less and collect that induced fluorescence over a time interval
of 50 if
not 100 ns starting at the arrival of the UV excitation pulse on the probe
surface.
As the technologies associated with fast detection electronics, data transfer
and
computing speed progress over the next few years, it is anticipated that the
spectrometric and time domain detection subassemblies described above will be
combined by having each pixel of the linear array sensor at the exit imaging
plane of
the spectrometer collecting the time decays associated with the fluorescence
measured within each resolved spectral intervals simultaneously.
Such approach is expected to improve significantly the classification power of
the
resulting instrument by applying the multivariate analysis on an ensemble of
samples each having 2 dimensions (induced fluorescence time decay signatures
for
CA 02796489 2012-11-22
each resolved spectral intervals) instead of 1 dimension (induced fluorescence
time
decay for the bulk fluorescence and the spectral signature of that
fluorescence).
The computer 28 is housed in the electronic enclosure and controls the
different
functions of the system. It controls the user interface provided by the
electronic
display 14, which is equipped with a touch screen and carries several of the
user
interface functions. The computer 28 allows configuring and controlling the
excitation light source unit 22, the spectrometric subassembly, the time decay
detection subassembly and the camera 40 to achieve the synchronized
acquisition
process during a sweep or during still image acquisitions. The computer 28
also
manages the data transfers received from the spectrometric subassembly, the
time
decay detection subassembly and the camera as well as their recording. It
controls
the autofocus function of the objective based on the images provided by the
camera
during a sweep or during still imaging acquisitions. It performs the
multivariate
analysis on collected data ensembles. It records and presents the results of
multivariate analysis to the user. It provides minimal library field
management
functions for the user. It controls the data ports 50 with external computer
systems
to backup the acquired data and receive library and program maintenance
updates.
The computer 28 runs two main algorithms, one for the spectral domain and
another for the time domain of the fluorescence signal. These are designed to
process the raw data collected by the system. Each algorithm is designed with
considerations based on a fluorescent signal models in both the spectral and
time
domains. The two dimensions are decoupled. The spectral signal is integrated
over a
time interval much greater than the pulse duration and the time signal is
spectrally
16
CA 02796489 2012-11-22
integrated. Signatures exist depending on the material for the spectral domain
and
the time domain. The signature in the time domain is modeled as an exponential
decay. The two signatures contribute to the classification of the material at
the
origin of the fluorescence. For each probed surface element, the data consists
in two
vectors: one is a function of time and the other is a function of the
wavelength. All
fluorescent materials have a spectral and a time response that is
characteristic of
their class. The basic signal can be modeled by the following equation:
s(A.,t). b(A.,t)+c(il,t)
The fluorescent signal induced by the UV light pulse is s. b and c are the
background
surface and the trace contaminant fluorescing signals, respectively. With the
uses of
a beam splitter, system splits s in two parts more or less equals (spectrally
and
dynamically). The first part is directed to the time decay detection
subassembly and
the other part is directed to the spectrometric subassembly. During the
detection
process, noise is added by the detection electronics to the intrinsic photon
shot
noise associated with the particulate nature of light. However, it is assumed
that the
electronic noise is negligible in comparison of the photon shot noise. The
detected
spectral signal is defined as
T
SR)= .1" fn(A)S(A ,t)dtd
where s(2.,, ) is the detected spectral signal in the band n. (A) is the
sampling
function of wavelength for each band and depends on the sensor spectral
response
17
CA 02796489 2012-11-22
characteristics. T is the duration during which the fluorescence signal is
integrated
by the spectrometric subassembly.
For the time signal, the problem is different since there is a time response
associated with the excitation. The UV light pulse pumps molecules of the
irradiated
materials to an excited level. Then, these excited molecules go back to their
ground
(unexcited) levels after random time periods. The number of excited molecules
decreases exponentially since there is no dependency between the excited
materials.
Since the number of excited molecules decaying in a given time interval is
proportional to the total number of excited molecules, the fluorescent signal
decreases following an exponential function. The dominating photon shot noise
is
modeled as a Poisson distribution. The acquisition system is modeled by a
linear
filter followed by an analog to digital converter. The resulting digitized
signal is
modeled using the following equation where sLdescribed the temporal
characteristics of the UV light pulse.
tfl t Amax
s(t) = f h(t ¨ T, A) SL(t) da dr dt
tn_i 0 Amin
In this equation, n denotes the nth time interval, TiS the convolution
variable and
h(t ,A,) is the impulse response of the system composed of the detection
electronics
and of the UV light pulse dynamic shape. The output signal will be composed of
a
rising portion followed by a maximum and an exponential decay afterward. Since
the system is linear and lowpass with a cutoff frequency higher than the
frequency
contained in the signal, it will follow the exponential decay of the
contaminants with
18
CA 02796489 2012-11-22
a delay. Therefore, the signal processing for this process consists in
estimating the
relaxation or time constant of a decreasing exponential buried in electronic
and shot
noise. The relaxation constant will provide information related to the nature
of the
fluorescing molecules contained in the contaminant.
The fluorescence is acquired at the same time in the spectral and time
domains.
Once the spectral processing has been performed and a given pixel is qualified
as a
spectral anomaly, the relaxation constant is estimated from the time domain
fluorescence signal. The resulting relaxation constant is appended to the
spectral
information to further refine the classification of the fluorescing material
The algorithm designed to process the spectral data acquired by is designed
with
the following assumptions:
1. The surface probed to acquire a single fluorescent data set (spectral and
dynamic)
is very small.
2. The background surface associated with each probe is the same.
3. Many probes can be acquired to assess the entire swept surface.
4. There is sufficient contrast between the fluorescence spectra of the
background
surface and the contaminant.
The algorithm for processing the data performs the following steps:
1. Dataset acquisition (A statistically significant number of spectral/dynamic
vectors)
2. Data pre-processing (Spectral binning)
19
CA 02796489 2012-11-22
3. Data classification
a. Background
b. Not background
4. Background statistic computation
a. Mean
b. Covariance matrix
5. On all spectral vectors compute the two following detectors
a. Anomaly detection
b. Match filter on library of signature
c. Outcome may be one of the following
1. Uncontaminated
II. Contaminated unknown
III. Contaminated known
i. Hazardous biological material
ii. Hazardous organic material
iii. Energetic material
iv. Benign biological material
v. Benign organic material
6. Categorization
CA 02796489 2012-11-22
a. Analysis of the data for output to the reporting step
7. Reporting
1. The dataset acquisition ensures that the collected vectors are in a
sufficient
number to ensure that the statistics that will be computed will be valid and
represent the effective values of the statistics of the background when
about half of the vectors are used if the vectors are from an uncontaminated
background surface. All the vectors are tagged and their location on the
probed
surface is known.
2. Data pre-processing is used to reduce the number of spectral bins in the
spectral
band measured by the sensor to a number that is sufficient to represent
adequately
the fluorescence phenomenology.
3. Data classification provides the identification of spectral vectors that
are
members of the background. This identification does not need to be fully
comprehensive in a first step. The complete identification of the spectral
vectors
associated with the background surface is done in two steps. The first step is
the
computation of the background vectors direction. It uses a SVD (Singular value
decomposition) principal component analysis followed by a distance computation
of
each vector from the mean of the full set of data in the SVD space using the
first two
singular vectors. The second step uses a spectral angle mapper algorithm to
identify
all the vectors part of the background surface.
4. Background statistics computation
a. Mean of the background data subset.
21
CA 02796489 2012-11-22
b. Covariance of the background data subset.
5. Computation of detectors (performed on each acquired vectors)
a. Anomaly detection: the anomaly detection is performed on each vectors of
the dataset. The values obtained from the background vectors are used to
estimate the threshold to be used to characterize a given vector as a spectral
anomaly. The mathematical operator is the Square Mahalanobis distance.
b. Match filter: the match filter computes using a library of signatures. This
step is evaluated on the background vectors to evaluate the thresholds for
the detectors. When a detector output is higher than the thresholds for many
signatures value, the signature producing the best score above a given
threshold will be selected as the potential contaminant. If the vectors only
trigger an anomaly without having a sufficient score on a contaminant it will
be qualified as unknown.
6. Categorization
a. Each vector will be processed according to the data that have been
computed in the following set of categories
1. Background
The anomaly detector and the match filters do not have triggered in
any way on these vectors.
II. Spectral anomaly
22
CA 02796489 2012-11-22
I. Unclassified (the material cannot be sorted in a known
category): The anomaly detector result provides a value that is
above the threshold for anomaly detection. However the match
filters do not provide a classification level sufficient to tell what
the surface giving that vector is made of.
2. Classified
a. Hazardous biological material
The contaminant is reported as a biological compound
that could be harmful and could be considered as a
biological agent. The anomaly detector and the
signature identification provide sufficient certainty to
class the contaminant on the surface.
b. Hazardous organic material
The contaminant on the surface is organic in nature and
can be toxic or dangerous. The quantity of material may
be insufficient to be dangerous. However, the probed
surface has been in contact with a dangerous material
organic in nature. The anomaly detector and the
signature identification provide sufficient certainty to
class the contaminant on the surface.
c. Energetic material
23
CA 02796489 2012-11-22
The contaminant on the surface is reported as an
explosive or a precursor of an explosive. The quantity
may not be sufficient to show any risk. The anomaly
detector and the signature identification provide
sufficient certainty to class the contaminant on the
surface.
d. Benign biological material
The contaminant is reported as a benign biological
material. The anomaly detector and the signature
identification provide sufficient certainty to class the
contaminant on the surface.
e. Benign organic material
The contaminant is reported as organic, but not
dangerous. The anomaly detector and the signature
identification provide sufficient certainty to class the
contaminant on the surface.
7. Reporting
The report for the probed surface includes the swept area, its location on the
surface if it is possible and the proportions of vectors identified as
background spectral anomaly, as classified biological and as classified non
biological material. Based on these information and the amplitudes of the
24
CA 02796489 2012-11-22
collected signals, an evaluation of the quantity of contaminants per surface
area is derived.
For each time decay vectors, four signal zones are identified: the rise, the
transition,
the relaxation or decrease, and, the low signal zone. The relaxation zone
contains
the information about the fluorescence time decay, which can be derived from
the
slope of the logarithm of the amplitude in that portion of the signal. In the
rise and
transition zones, the induced fluorescence reacts to the UV light pulse shape
and the
linear signal processing acquisition system. If only the relaxation portion is
used,
assuming there is no excitation at that time and the acquisition system is
fast
enough to react to the decreasing exponential, the only remaining contribution
to
the signal time decay is defined by this logarithmic slope. As a result, the
information is simple to extract in that signal zone. In the last signal zone,
the
information content is such that noise overwhelms it and including it in the
process
will decrease the accuracy of the relaxation time derivation. Figure 2 shows
an
example of a transient fluorescent signal. The first section where the signal
is rising
is from 0 to about 2 ns. The signal then decreases slowly until approximately
14 ns.
After 14 ns, instrument noise prevails as shown in Figure 2.
The process designed to extract the relaxation constant of the fluorescence
should
contain the following components:
1. Determination of the beginning of the exponential decay zone;
2. Determination of the useful signal zone;
3. Estimation of the amplitude of the exponential decay;
CA 02796489 2012-11-22
4. Estimation of the relaxation constant.
The beginning of the decay zone starts when there are no more excitations
provided
by the incident light pulse and after a time interval related to the response
time of
the detection electronics. The knowledge of the UV pulse duration and of the
bandwidth of the detection electronics will contribute in determining the time
delay
before entering the decay zone. Assuming the detection electronic response is
negligible in comparison of the UV pulse duration, the first sample of the
decay zone
should be located at least at two-pulse width of the peak of the collected
signal. This
peak is the maximum of the fluorescence time function.
As the time from the beginning of the decay zone increases, the signal to
noise ratio
decreases. At some point, there will be no increase in information by adding
one
more data value. Adding points beyond that point will diminish the quality of
the
estimation. The determination of this last exploitable sample is performed
with the
use of the peak value. The choice is made using the following procedure.
1) Estimate the noise of the instrumentation using points that are very far
from the location where the signal is located;
2) Multiply by 2 this value and set it as the threshold;
3) As the signal decays, detect the first point that is below that threshold;
4) The last point included in the set of points for the estimation of the time
constant is the point before the one detected in 3;
5) If there is no such point then use the first two points after the peak
value.
26
CA 02796489 2012-11-22
Two parameters have to be estimated. The more important is the time constant
which is the rate at which the fluorescence is decaying. The second is the
amplitude
of the exponential. Two methods for this estimation can be used. The first
method is
based on the LMS (Least Mean Square) criterion in which the sum of the squares
of
the errors is minimized as a function of the two required parameters. This
method
yields the following equations.
I A ex
In (s(t,7))1it,72 ¨Etn ¨Nit,, ln (s(t,, ))
p=
NEtn2 ¨ (Et,, )2
a = Eln(s(t,i))Itn ¨ NEtn in(s(tn))
NIt,72 ¨ (1,02
Another way of estimating the parameters is based on the maximum likelihood
technique, where the likelihood of a value is maximized. In this last case,
the method
requires the assumption of a probability distribution to represent the
likelihood. To
provide the estimators, the Poisson distribution is used because it is the
model for
shot noise, which is the major contribution in the signal.
The following equations are used as the basis of the computation. Since these
two
equations cannot be solved directly, minimization techniques or zero finding
iterative methods are used. The amplitude of the exponential requires the
estimation of the time decay.
For the maximum likelihood estimation, the estimators are given by
Etnexp(¨at,) Etns(tn)
and
Iexp(¨at) s(t)
27
CA 02796489 2012-11-22
A =,(tn)
Lexp(¨at)
The extraction of the time decay constant from the maximum likelihood
estimation
is performed using a least square error, or finding the zero of the following
function:
F
exp(¨at) s(t
(a )= _______________________________
Iexp(¨w) Is(t)
or
a min
Etn exp(¨atõ Etfis(tõ))2
:
Iexp(¨can) Is(t)
This process is performed for each pixel where fluorescence is detected. The
estimated relaxation constant is appended to the spectral fluorescence
information.
It will thus be seen that embodiments of the invention provide a convenient
tool for
first responders to recognize the presence of potentially hazardous
substances,
which is rapid and portable.
28
