Methods and systems for signal processing in particle detection systems

• US 7,242,009 B1
• Filed: 06/22/2005
• Issued: 07/10/2007
• Est. Priority Date: 06/22/2005
• Status: Active Grant

• Current Assignee: Hach Company (Veralto Corp.)
• Original Assignee: Hach Ultra Analytics Incorporated (Danaher Corporation)
• Inventors: Wilson, Geoffrey, Brady, James
• Primary Examiner(s): Porta; David
• Assistant Examiner(s): WEBB, CHRISTOPHER G
• Application Number:
  US11/158,491
• Time in Patent Office: 748 Days
• Field of Search: 2504581-4611
• US Class Current: 250/461.1
• CPC Class Codes: G01N 1/2273 Atmospheric sampling
• CPC Class Codes: G01N 15/1456 without spatial resolution of the texture or inner structure of the particle, e.g. processing of pulse signals
• CPC Class Codes: G01N 2001/022 sampling for security purposes, e.g. contraband, warfare agents
• CPC Class Codes: G01N 2001/025 postal items
• CPC Class Codes: G01N 21/53 within a flowing fluid, e.g. smoke
• CPC Class Codes: G01N 21/6402 Atomic fluorescence; Laser induced fluorescence
• CPC Class Codes: G01N 21/6486 Measuring fluorescence of biological material, e.g. DNA, RNA, cells G01N21/6428 takes precedence

Abstract

A method for statistically distinguishing between two or more populations, such as between samples containing background and threat particles, is disclosed. According to one embodiment, the method for statistically distinguishing is a method for processing a signal in a particle detection system. The method may include characterizing each particle detected by a signal vector. The method may further include calculating a weight of each particle detected using a weighting function of the signal vectors. The weighting function may be a functional of a threat particle probability distribution and a background particle probability distribution. The method may also include summing a total weight of all particles tested and generating an alarm signal if the total weight exceeds an alarm level.

Claims

1. A method of using laser-induced fluorescence to distinguish (a) a presence of threat particles interspersed with background particles from (b) an absence of threat particles, the threat particles representing fluorescent biological substances, comprising:

using a multiple-wavelength detection-band laser-induced fluorescence sensor to acquire a measurement set of particles for laser-induced fluorescence analysis, the sensor providing threat-present and threat-absent particle signals in response to the acquired measurement set of particles;

defining a grey-scale weighting function formulated on competing sensitivity and false positive rate factors, the sensitivity factor representing sensitivity of detecting threat particles against non-threat particles and the false positive rate factor representing a probability of falsely concluding that the acquired measurement set of particles contains threat particles;

using the weighting function to assign weights to particles in the measurement set, based on a numerical characterization of a particle and a weight associated with that characterization;

assigning weights to each particle in the measurement set, based on the weighting function; and

processing the weights assigned to the particles and producing an indication whenever a result obtained from the processing of the weights represents a likelihood of concluding correctly that the measurement set of particles contains threat particles.

2. The method of claim 1, in which the threat-present and threat-absent particle signal distributions exhibit an overlapping region, further comprising optimizing the weighting function to diminish the overlapping region.

3. The method of claim 1, in which the processing includes accumulating the weights assigned to the particles, and in which the result of the process is an accumulated weight that produces the indication.

4. The method of claim 3, in which the threat-present and threat-absent particle signal distributions exhibit an overlapping sum of weights distributions, wherein the method further comprises optimizing the weighting function by substantially minimizing an overlap integral of the threat-present and threat-absent sum of weights distributions.

5. The method of claim 3, in which the weighting function is derived from a pre-determined fluorescence-space distribution of threat particles and from a distribution of background particles that adapts to a changing background.

6. A method of deciding whether a particle-containing sample of an air stream, which air stream may contain particles of interest and background particles, is a sample of interest, comprising:

detecting a measurement set of particles in the air stream;

defining a numerical characterization scheme for characterizing particles in the measurement set;

individually assigning a numerical characterization to each detected particle in the measurement set;

defining a grey-scale weighting function, with a range of more than two weights, capable of associating a weight with a numerical characterization;

using the weighting function in assigning weights to particles in the measurement set, based on the numerical characterization of a particle and the weight associated with that characterization;

processing the assigned weights for detected particles in a sample; and

producing an indication, based on the processing of the assigned weights, whether the measurement set contains some particles of interest or background particles.

7. The method of claim 6 in which:

the indication has (a) a sample-of-interest distribution which characterizes the distribution of processed particle weights of samples of interest; and

(b) a background distribution of which characterizes the distribution of processed particle weights of background samples; and

wherein the weighting function is a function of the sample-of-interest distribution and a function of the background distribution.

8. The method of claim 7 in which:

the sample-of-interest distribution has an overlap with the background distribution; and

the weighting function assigns weights to reduce the overlap.

9. The method of claim 8 in which:

the sample-of-interest distribution has a mean and a variance;

the background distribution has a mean and a variance; and

the weighting function assigns weights to substantially maximize (a) the squared difference of the means of (i) the sample-of-interest distribution and (ii) the background distribution, divided by (b) a sum of the variances of (i) the sample-of-interest distribution and (ii) the background distribution.

10. The method of claim 9 in which:

the difference of the means of (i) the sample-of-interest distribution and (ii) the background distribution is held constant by exploiting an invariance of results to multiplicative scaling of the weighting function; and

the weighting function assigns weights to substantially minimize the sum of the variances of (a) the sample-of-interest distribution and (b) the background distribution.

11. The method of claim 10 in which:

minimizing the sum of the variances of (a) the sample-of-interest distribution and (b) the background distribution is performed using Lagrange'"'"'s method of undetermined multipliers.

12. The method of claim 6 in which:

the weighting function is based on a particle-weight distribution of background particles and on a particle-weight distribution of particles of interest.

13. The method of claim 12 in which the particles of interest are aerosolizable pathogen particles.

14. A device for analyzing samples of an air stream containing particles, comprising:

an excitation wavelength;

an emission wavelength that is different from the excitation wavelength;

an illumination area through which the air stream passes;

a light source, comprising light at the excitation wavelength, that illuminates, in the illumination area, the particles contained in the air stream;

an excitation light detector that detects light from the particles at the excitation wavelength;

a fluorescence detector that detects light at the emission wavelength;

a memory that accumulates particle information, including (a) data representing a count of a number of particles that has passed through the illumination area and (b) data representing light detected by the fluorescence detector;

a sample-size criterion comprising data representing a maximum number of particles to be included in a complete sample, the complete sample comprising a plurality of particles; and

a sample-completion testing module that applies the sample-size criterion to determine when the memory has accumulated particle data from a complete sample.

15. The device of claim 14 in which the sample-size criterion further comprises information representative of a time interval.

16. The device of claim 14 in which the maximum number of particles to be included in a complete sample is predetermined.

17. The device of claim 14 in which the particle information accumulated in the memory comprises data representative of light from the particles that has been detected by the excitation detector.

18. A system for reducing the false positives in a particle detection system, comprising:

a laser that irradiates a particle-containing airstream containing unknown particles which may be particles of interest or background particles;

a fluorescence detector that detects fluorescence emitted by the particles in the airstream in response to irradiation by the laser;

a non-fluorescence detector that detects particles in the airstream regardless of fluorescence;

a computer-readable medium for storing;

a sample size, expressible or convertible into in time, to determine a size of a sample of detected particles a sample volume rate expressible in liters per second;

a sensitivity expressible by a number of Bacillus globigii particles amid a number of total particles per volume;

a mean time between false positives; and

a true positive confidence level;

where a common log of the mean time in seconds between false positives is greater than 4 for a true positive confidence level of 90% and a sample time, sensitivity, and true positive confidence level for conditions with ratios of;

(a) 10 seconds, to (b) a ratio of 42 Bacillus globigii particles per 2500 total particles per liter.

19. The system of claim 18 in which the common log of the mean time between false positives is greater than 5.

20. The system of claim 18 in which the common log of the mean time between false positives is greater than 6.

21. The system of claim 18 in which the common log of the mean time between false positives is greater than 7.