GB2368390A

GB2368390A - Analysing a plurality of objects such as particles in a fluid

Info

Publication number: GB2368390A
Application number: GB0006381A
Authority: GB
Inventors: Peter Smith
Original assignee: Loughborough University; Loughborough University Innovations Ltd
Current assignee: Loughborough University; Loughborough University Innovations Ltd
Priority date: 2000-03-16
Filing date: 2000-03-16
Publication date: 2002-05-01
Anticipated expiration: 2020-03-16
Also published as: GB2368390B; GB0006381D0

Abstract

A plurality of objects (17) are analysed by measuring the intensity of radiation modified by the objects in a radiation field (15) including a background intensity, eg by scattering or obstruction of optical or acoustic radiation. The mean and variance of the intensity measurements are calculated for at least two instances distinguished by time, position, source wavelength, or polarisation. A processor computes the ratio of the squared difference of the means to the difference in the variances. This ratio, or a function of it, is used to determine the number or type of the objects. Light from a source (13) may be scattered by particles in a stream 19 to two detectors (25,27) to produce measurements at different positions (21,23) or measurements at one position at different times. Alternatively measurements may be made at different scatter angles, eg to determine the number of particles of one size in the presence of particles of another size.

Description

Analysing Data

Field of the invention The present invention relates to methods for extracting information from observed data, to computing devices arranged to perform the methods, and to computer program products carrying program instructions for performing the methods.

Background of the invention The majority of object counting methodologies are based on an imaging approach (e. g. microscopy for erythrocyte concentration) or on the modulation of a radiation source by the passage of objects through the radiation field (e. g. particle counters via obscuration or scattering). An essential feature of these methods is that objects are individually counted by a separation in time or space. There are numerous examples of art which apply these basic ideas in a range of applications. The following list is representative of this technology : *) In-vitro cell analysis [e. g. US Patent Nos. 5703959, 5274431] * Airborne contaminants [e. g. US Patent Nos. 5467188,5319575, 4940327] * Surface inspection (e. g. semiconductors) [US Patent No. 5311598] Smoke/dust detection and discrimination [e. g. US Patent Nos.

5670947, 5581241] 'Contamination of power station boiler condensate [e. g. US Patent No. 4672216] * Contamination of injection moulded plastic resins [e. g. US Patent No. 5684583] * Identification of targets in radar clutter [e. g. US Patent No. 5612700] 'Fuel contamination [e. g. US Patent No. 5200064] . Engine oil soot contamination [e. g. US Patent No. 5309213]

Rainfall detection and differentiation (rain/sleet/snow) [e. g. US Patent No. 5557040] Other more heuristic approaches to particle suspension analysis are based on turbidity, nephelometry, spectroscopy and combinations thereof.

These heuristic methods yield surrogate measures of the objects comprising the suspension.

When the interaction between objects and a radiation field is dynamic, either due to a dynamic of the objects themselves or due to a translation between the source and an extended object field, the obscured or scattered field is also dynamic. The methodologies which use measures of the fluctuating radiation field to characterise the objects are those closest to the present invention and may be categorised as follows 1. Dynamic light scattering (small particles undergoing Brownian motion) [e. g. US Patent No. 5434667] 2. Discrimination via the signal to noise ratio [e. g. aerosol discriminator: US Patent No. 5293049] 3. Discrimination via measures of mean, variance and temporal correlation [e. g. analysis of flocculation content: US Patent No.

5194921] 4. Signal identification in non-Gaussian background clutter [e. g. US Patent No. 5694342] In dynamic light scattering, the objects under investigation are assumed to be in Brownian motion and therefore their dynamics (temporal autocorrelation) can be related to their kinetic energy and hence the mass. The assumed motion is only true for small particles (sub-micron) in liquid or gaseous suspension. The number of particles is not measured.

In discrimination via the signal to noise ratio, the signal to noise ratio (S/N) of optical intensity fluctuations is used to discriminate between airborne particles

of various types. An empirical relationship between the particle type and S/N is suggested. The number of particles is not measured. The method is claimed to be able to discriminate between smoke particles (high S/N) and larger particles such as dust and fog (low S/N).

In discrimination via measures of mean, variance and temporal correlation, a mixture of particles in a floculation process is analysed by measurement of the mean and variance of optical intensity fluctuations arising from two sources of contrasting wavelengths, together with the correlation coefficient between the two signals. These measures are entered into a formula which calculates the concentration of the dissolved coagulant. As the process proceeds the amount of dissolved coagulant decreases and hence the progress of floculation can be monitored. The number of particles is not measured.

In techniques of signal identification in non-Gaussian background clutter, an analysis of a coherent imaging system (such as microwave radar) is performed in which the noise is divided into Gaussian and non-Gaussian components so as to extract a signal embedded in the non-Gaussian component. A measure of the signal to noise ratio is used to decide whether a signal is present or not. The method does not count or characterise targets by this statistical method.

Summary of the invention The present invention attempts to provide new and useful methods for inferring the number and/or type of first objects (A) in the presence of a background (caused by second objects B).

This invention addresses a generic problem in multi-object analysis : namely the detection and characterisation of objects (or other centres) in the presence of an unspecified background.

Within the scope of the invention at its most general level, the background objects B may be of any (e. g. unknown) character or characters,

and of any number. Objects A and B may also be interpreted as representing any form of material inhomogeneity, e. g. turbulence. The method employs at least one source of radiation (e. g. electromagnetic (e. g. optical) or acoustic) and at least one detector. Each detector is assumed to measure a portion of the radiation scattered from (or obscured by) objects moving through the radiation field of the source. Both objects A and B may be fluctuating in number and may be randomly positioned. The mean and variance of the intensity fluctuations are calculated for at least two instances distinguished by time, position, source wavelength or polarisation. A processing element computes the ratio of the squared difference in the mean and the difference in the variance. This ratio (or a function of it) is used, in various embodiments, to determine the number and/or characteristics of the objects A.

As discussed below in detail, under certain conditions this ratio is independent of the absolute measurement. In principle therefore any embodiment or apparatus can be used within the scope of the invention, provided that the fluctuations are measurable. The simplicity of the calculation used to obtain the particle number or type is appropriate for low cost implementation, for example, and is thus appropriate in consumer markets.

Applications exist in the measurement of small objects such as those present in fluid suspensions (e. g. contaminated water, milk, paints, food stuffs) and gaseous particle suspensions (e. g. contaminated air, rainfall, pollen, fog, soot, pharmaceutical powders) or in such other diverse fields as the monitoring of living creatures (e. g. bacteria, shoals of fish), biological suspensions (e. g. cells, erythrocytes, viruses) and artefacts (e. g. projectiles).

It is also possible to use the invention to analyse any form of material inhomogeneity for which the object numbers are replaced by effective numbers which are characteristic of the inhomogeneity. Example applications include material surface inspection, biological tissue analysis and remote sensing.

Specifically, a first expression of the present invention is a method of using measurements of the intensity of radiation scattered (or obscured) by a population of scattering (or obscuration) centres in a radiation field, the

measured intensity including a background intensity, to determine at least one characteristic of the scattering centres, the method comprising the steps of : measuring a first sample set of the radiation scattered (or at least partially obscured) by the centres; measuring a second sample set distinct from the first sample set; computing the mean (I,) and variance a, of the first sample set, and the mean < > and variance a, of the second sample set; and computing a ratio

Rv =A ( (I,)-BI,) y (1) a,'-ca.'' Y-2-CO'2

where A, B and C are constants. Alternatively any function of Rij may be calculated (e. g. a trivial function such as Rij plus a constant, or any more complex function).

For simplicity, the preferred features of the invention will be explained below with reference to scattering only, but they are applicable also to the case of obscuration. Indeed, the first and second sample sets may in principle differ in that one uses scattered radiation and the other is of at least partially obscured radiation.

The scattering centres may be first objects present in the radiation field, but alternatively they may represent any other inhomogeneity, (e. g. a tissue inhomogeneity) The background intensity may be due to scattering, and may for example be caused by a set of second objects in the radiation field. That is, the first objects represent only a subgroup of the total number of objects in the radiation field.

The first and second sample sets may differ for example in any one or more of the following ways: being measured at different times, being measurements of different wavelengths in the radiation, being measurements of different radiation polarisations, being measurements from different angles in relation to the field, or being measurements scattered from different positions in the radiation field.

In some embodiments, the centres are moving relative to the radiation field (either of both of the field and centres may be moving in absolute terms).

For example, in some embodiments, all the objects are flowing through the field, and the two sample sets may be made in respective locations along the direction of flow.

We observe here that the radiation field need not be contiguous: rather it may include two or more spatially separated regions in which there is radiation, and in such a case, if the two sample sets differ in position those two positions may be in different respective regions.

As explained below in detail, if a subgroup of the object population has changed in number, Ry may indicate the change; if a subgroup has changed in nature, then R may indicate this.

The technique is successful in cases in which the signal from the background varies little between the two sample sets; for example, if the two sample sets are taken at times differing by to, while time variation in the background occurs only on a much longer timescale (e. g. 10 or 100 times longer). The technique is particularly successful in cases in which the statistics of the background population have a low contrast factor with respect to the sample sets.

Optionally, one or more further sample sets can be derived, and used.

For example, if a sub group of the object population has changed in both nature and number, at least one further set may be used to compute the ratio Ry/R'k (where i, j and k indicate different respective sets of samples) which is a characteristic of the property of the changed sub group and is independent of its number.

The constants A, B and C may be selected in accordance with the situation in which the invention is applied, as explained in more detail below.

In many embodiments, the constants B and C are equal to one, and constant

A may also be 1, so that

Ry 2 .--- (2)

Although the invention has been expressed above as a method, the invention further relates to an apparatus arranged to perform the method, and to a computer program product, such as a recording medium, carrying program instructions readable by a computing device to cause the computing device to perform the computing steps of a method according to the invention.

Theoretical explanation An incoherent radiation field is assumed to illuminate an arbitrary population of objects, randomly distributed in space and moving through a measurement volume. Assuming a one-dimensional pattern, the total scattered (or obscured) intensity from M groups of Nk independent objects may be represented by

in terms of the intensity q. It will be assumed that objects are distributed in space according to a uniform random variable suitably normalised and that the number of objects of a given category may fluctuate independently of their scattering characteristics or position. The first two intensity moments will be

On assuming the particle populations all vary according to Poisson statistics and noting that

equations (4) and (5) can be used to form the normalise moment

We now turn our attention to the engineering problem of recognising a population (A) of first objects in an arbitrary background (B). If the objects A are to be observed by statistical means, they must add significantly to the background fluctuations, although it is stressed that the total background fluctuations may still exceed those generated by objects A. A version of equation (6), may therefore be written

Now assume that the situation in question is such that change in the total fluctuations are assumed to be due to the objects A alone. Such changes may be a function of one or more of the following independent variables . time position (e. g. scattering angle)

o wavelength o polarisation (e. g. vertical or horizontal linear polarised laser) Instances of equation (7) and the corresponding mean may then be represented by

It will also be useful to use a representation in terms of the variance crl

By considering instances of equation (8) and (9) we are now able to eliminate the background fluctuations and the coupling constant by forming the ratio

The quantity Ry, which is independent of the intensity Io, is itself an interesting characteristic. However, further progress can be made in two physically important cases

(NAt) (qAi-qAj Y . Case (i) : (NA) = (N) = (N) R = (NA,) (qAI-qAl) Ai Ai (11)

10 . Case (ii) q, === (NA,)- () (12)

This example also demonstrates why any function of Ry, even a trivial function thereof (such as its square), may be of interest.

In Case (i), a third measurement may be performed to obtain a characteristic of the objects A independent of their number:

In practice a process of calibration may be performed to overcome any problems of the non-ideal performance of the measurement system or due to a relaxation of the requirement for the objects B to create the same fluctuations in all instances. Thus the preferred theoretical formulae to be used are

Where A, B, C, D, E are determined by calibration in the particular embodiment.

Brief description of the figures Non-limiting embodiments of the invention will now be described, for the sake of example only, with reference to the accompanying drawings, in which: Figure 1 (a)- (b) shows two arrangements in which radiation is scattered by objects within a radiation field, and two measurement sample sets are taken;

Figure 2 shows scattered radiation intensity as a function of angle for spherical particles in the Mie theory ; Figure 3 shows the results of a simulation of a method according to the invention; Figure 4 shows measured intensity (vertical) as a function of time (vertical) in a experimental test of a method according to the invention; and Figure 5 shows the experimental results derived by the method according to the invention from the data of Figure 4.

Embodiments of the Invention Figure 1 (a) shows schematically a first arrangement in which the invention may be employed. A radiation field 1, produced by a source 3 and regulator 5 is scattered by objects 7. Radiation 9,11 propagating in two directions is measured by respective measuring devices.

In the arrangement of Fig. 1 (a) the sample sets may be separated by time or by a characteristic of the objects which induces differential fluctuations in the radiation field (e. g. differential absorption at two separate wavelengths).

Figure 1 (b) shows schematically a second arrangement in which the invention may be employed. A radiation source 13 produces radiation which is spread by device 14 to produce a radiation field 15. Objects 17 flow along a dotted flow path 19 through the radiation field and the intensity of radiation scattered in two regions 21,23 is measured by measuring devices 25,27.

Note that the device 14 may alternatively be a beam splitter, to produce two beams, each of which is fed to a respective one of the regions 21,23, or alternatively two beams may be generated by two respective radiation sources. Thus, the objects 17 only encounter radiation in the two regions 21, 23 which, at least as viewed along the flow path 19, are not contiguous.

In the arrangement of Fig. 1 (b) the sample sets are measured at different respective positions, and, although the two sample sets may in fact be taken at the same time, from the point of view of the objects they are at different times, in the sense that if objects evolve between the two regions 21, 23 then the two measurements will capture different stages of their evolution.

Optionally, the sample set may differ also in any one or more of the factors mentioned above in relation to Fig. 1 (a).

Optionally, if the flow itself is evolving, then the timing of the measurement may reflect this (e. g. such that an evolution of the background does not make it impossible to observe the foreground). For example, the measurements may be taken at respective times spaced apart by the time it takes objects to travel between regions 21,23, so that the same objects will be observed at the two positions.

In any embodiment of the invention, each sample set is produced by taking multiple measurements, preferably spaced apart by short time intervals. For example, each sample set may include at least 100 measurements, at least 1000 measurements or at least 10000 measurements, depending upon the required dynamic range of measurement.

In the case that the first and second sample sets are distinguished by time to, the respective pairwise time intervals between the plurality of measurements in each sample set are preferably much less than to, e. g. at least 50 times smaller or more preferably at least 100 times smaller.

We have identified at least eight embodiments based on the cases (i) or (ii) and the four differential contrast factors (i. e. differences between sample sets): time, position, wavelength and polarisation. These contrast factors are used in various embodiments of existing art but not in conjunction with the statistical analysis described here. The nature of each embodiment is appropriate to a different generic application and these are listed in the table

below together with illustrative example applications.

Case Contrast Generic Example Industry Factor Application Application (i) time Evolution of object Making particles pharmaceutical type (e. g. crystals) (i) position Object identification Particle discrimination consumer (e. g. soot, dust) (i) wavelength Object identification Particle discrimination Healthcare (e. g. erythrocytes)

polarisation Object identification Particle shape Environmental discrimination & healthcare (e. g. asbestos) (ii) time Evolution of object Contamination Food, water, number monitoring semiconductor (e. g. filter breakthrough) (ii) position Object mapping Particle distribution Oil & gas (e. g. pipeline monitoring) wavelength Multi-species Objects discriminated Healthcare counting by colour (e. g. cancer (same cross-vs normal cells) section) polarisation Multi-species Objects discriminated Environment counting by shape (same cross- (e. g. airborne pollen, section) fibres)

A particularly interesting example is case (ii) when the sample sets vary by position (as in Fig. 1 (b)). An example of this situation is measurements of the flow produced using a medicament dispenser apparatus, such as a dry powder inhaler, of a medicament (drug) carried on a carrier. In the flow which leaves the apparatus, the medicament is present on particles (objects) of the carrier, but gradually the medicament may become separated. Thus a measurement of a first sample set of the flow close to the device will have a background generated by the drug carrier, and no foreground (medicament) particles. A second sample set downstream will still have as a background the scattering caused by the carrier (this is largely unchanged), but it will have as a foreground the drug itself, now causing significant scattering due to its separation from the carrier.

The present invention thus makes it possible to provide monitoring of the amount of drug being dispensed (e. g. amount per second) using relatively

unsophisticated measurement devices. This would permit a failsafe inhaler monitoring system to be produced which employs components cheap enough to the included in inhalers and which may be more reliable than present mechanical failsafe devices.

We will now discuss in detail two representative embodiments of the invention.

1. Computer simulation of Case (ii) (with instances separated by time) A computer simulation was performed assuming a discrete population of independent particles having Poisson number fluctuations (since each species is independent of the others and since only individual particles can be in the beam of the incoming radiation at any given instance in time). This applies to both polydispersed background (B) particles as well as the monodispersion of interest (A). The background particles were assumed to have a range of sizes, modelled by a Gaussian distribution (having a mean particle size of 2.1 microns and a standard deviation of 0.5 microns). The A particles of interest for the current simulation were assumed to be 5 microns in size. All particles were assumed to be spherical and exact scattering intensity profiles were calculated by Mie theory. The incident radiation was assumed to be incoherent light of wavelength 633nm and both particle populations were assumed to have the same relative refractive index (m=1.2).

In the simulation, the sample sets were assumed to be distinguished by being at different scattering angles since the two particle populations have different scattering properties as a function of angle as shown in Figure 2.

Each simulation starts with the assumption that only background particles are initially present. First and second normalised moments were calculated for 500 sample populations at different angles and recorded. The five micron particles were subsequently added to the fluctuating background

at uniform steps (N AI) = 0, (N AJ) = 1, 2, K, 10. For each (N AJ) I 500 independent sample population of background as well as the 5 micron particles were generated and intensities were added incoherently for each angle. First and second moments were calculated and equation (15) was used (with

A=B=C=1) to determine the number of 5 micron particles in the total population. Figure 3 shows the calculated versus actual mean number of particles A for several forward and back scattered angles. It may be noted that the precision is not uniform with angle due to the large changes in scattering intensity with angle.

2. Experimental test of Case (ii) (with instances separated by time) Experiments were carried out using a flow system and intensity fluctuations were measured by a CCD camera imaging at far field. The sample volume was illuminated by a 30 mW He-Ne laser (=632. 8nm). The closed loop flow system was filled with deionized water at the beginning of experiment where flow was maintained by means of a peristaltic pump and particles were added while the system was operational.

At each instance of time, a frame was captured and the intensity probability distributions (paf) were calculated for an area within the image (40,000 8-bit pixel values). One hundred (i. e. sample size) of such pdfs were used to calculate the first and second moments for an initial frame-set (each frame-set contains 4 seconds of video at 25 frames/second) without any particles and henceforth for the remaining frame-sets. Figure 4 shows the change of pdf as a function of time. Particles (4.23 micron size) were added to the flow at regular intervals. Twenty two frame-sets were sampled at regular intervals for four consecutive injections of particle boluses into the flow. For each frame-set first and second moments were calculated and equation (12) was used to estimate the number of particles in the measurement volume.

This is done by taking the first sample set i as being the one at time=0, so that at all further times < N > is found by comparison to the time=0 measurement. In this case there are no objects as such in the radiation field when the first sample set is taken: the background is caused by scattering caused by turbulence in the fluid.

Figure 5 shows the results for the 22 instances of frame-sets. Initially there are no particles for two frame-sets. Particles are added to the reservoir at frame-set three and at regular intervals of five frame-sets. The plot shows as points connected by lines, the change of particle number calculated from measurements (using Equation (12) ) for each frame-set. For each injection, there is a delay of one frame-set before changes in fluctuations are measured due to the finite transport time to the measurement chamber. This is observed by a delayed response of estimated number. Following each injection, there is a big step as the high concentration of particles pass through the cell and this is followed by a gradual decay in particle number as the particles became dispersed throughout the volume. For purposes of comparison, Fig. 5 also includes a solid line which indicates the particle number estimated based on the particle concentration which was injected.

Note that it is possible to redefine the background at any moment to be the instantaneous level of signal. In other words, j would be reset to be the present instant, and equation (12) would then be used to derive the increase or decrease in the particle number compared to the present time.

Industrial applicability Among possible industrial applications are: Suspension analysis (counting and discrimination)-water quality, food processing & quality, process engineering, contaminant monitoring (e. g. filter breakthrough) * Aerosol analysis (counting and discrimination)-air quality, soot, exhaust, smoke Powder analysis (counting and discrimination)-pharmaceuticals Sedimentation analysis (e. g. blood cells) Cell counting & discrimination (e. g. cell division, labelling)

Claims

Claim

1. A method of using measurements of the intensity of radiation modified by a population of centres in a radiation field, the measured intensity including a background intensity, to determine at least one characteristic of the centres, the method comprising the steps of: measuring a first sample set of the radiation intensity; measuring a second sample set of the radiation intensity distinct from the first sample set; computing the mean (I,) and variance u ; of the first sample set, and the mean < > and variance 0"j2 of the second sample set; and computing a ratio

Ru =A ( (I,)-B (I,) y (1) (1',)

or a function of Rill where A, B and C are constants.

2. A method according to claim 1 in which the centres are scattering centres which scatter the radiation field, and the measurements are of scattered radiation.

3. A method according to claim 2 in which the centres are first objects present in the radiation field.

4. A method according to claim 3 in which the background intensity includes radiation scattered by a population of second objects in the radiation field.

5. A method according to claim 2, claim 3 or claim 4 in which the first and second sample sets differ in any one or more of the following ways: being measured at different times, being measurements of different wavelengths in the radiation, being measurements of different radiation polarisations, being measurements from different angles in relation to the field, or being measurements scattered from different positions in the radiation field.

6. A method according to claim 3 in which the first objects are moving into and/or out of the radiation field.

7. A method according to claim 6 in which the first objects are flowing with respect to the radiation field

8. A method according to claim 7 in which the two sample sets are measurements of radiation scattered in respective locations along the direction of flow.

9. A method according to claim 2 in which the scattering centres are inhomogeneities in a material present in the radiation field.

10. A method according to claim 1 in which the centres at least partially obscure the radiation, and the measurements are of at least partially obscured radiation.

11. A method according to any preceding claim in which one or more further sample sets are be measured, and used to compute a ratio /, or a function of this ratio, where k indicates one of said further sample sets.

12. A method according to any preceding claim in which the constants B and C are substantially equal to 1.

13. A method according to claim 3 in which the first objects are particles including a medicament.

14. A method according to claim 3 in which the characteristic of the first objects is their number.

15. An apparatus determining at least one characteristic of a population of radiation modification centres, and having: a radiation field generation system for generating a radiation field in which the centres are present; a measurement device for measuring a first sample set of the intensity of radiation modified by the population of centres and by a background; a second measurement device for measuring a second sample set, distinct from the first sample set, of the intensity of radiation modified by the population of centres and by the background; and a processor arranged to compute the mean (I,) and variance c, of the first sample set, the mean < > and variance of of the second sample set, a ratio

R =, 4 (I,)-B (I,) y (1) y J < or a function of R) j, where A, B and C are constants.

16. A pharmaceutical dispenser including a device for generating a supply of pharmaceutical including first objects which include a pharmaceutical, and an apparatus according to claim 10 to measure at least one characteristic of the first objects.