CA2148204A1 - Method and apparatus for cell counting and cell classification - Google Patents

Abstract

A method and an apparatus for analyzing a material within a container, such as blood within a capillary in a volumetric cytometry system provides for detecting the edges of the container, counting the cells within the container, characterizing the cells within the container, and evaluating channels of data which contain information relevant to more than one of the detectable characteristics of the cells. A scanner scans a container of material including certain cells. Sampling circuitry is coupled to the scanner to generate scanned images of the material in the container. Two or more scanned images are generated based on fluorescence data from dyes that have overlapping spectra. The two scanned images are processed using a linear regression analysis among corresponding pixels in the scanned images near certain cells to characterize relative contents of two fluorescing dyes in a target cell.
Target cells are identified from the scanned images using processing resources which identify a peak sample within a neighborhood, and compare the amplitude of the peak with the amplitude of pixels on the perimeter of the neighborhood. Upon identifying a target cell in this manner, data from the plurality of scanned images corresponding to the identified cell are saved for further analysis.

Description

'`

METHOD AND APPARATUS FOR CELL COUNTiNG
AND CELL CLASSIFICATION

~ ;l ., I llMlTFn COPYRIGHT WAIVFR
A portion of the d;sclosllre of this patent document contains material to which the claim of copyright protection is made. The copyright owner has no objection to the facsimile reproduction by any person of the patent document or the patent ~isclos~ ~re, as it appears in the U.S. Patent and Trademark Office file or records, but reserves atl other rights whatsoever.

CONTINUING APPl ICATION QATA

The present application is a continuation-in-part of Serial Number 08/018,762 filed February 17, 1993 en~itled "Method and Apparatus for Volumetric Capillary Cytometry", assigned to the same assignee of the present application.
CROSS-RFFFRFNCF TO RELATFn APPLI(`~
The present application is related to U.S. patent applicalion entitled "Method and Apparatus for Volumetric Capiliary Cytometry", invented by Baer et al, filed on the same day as the present application and owned by the same Assignee. The related application is incorporated by reference as if fully set forth herein.

AKornoy Dock-t No.: E' "~ ' ~H
mah/blT il2X2.X2 `_ BACKGROUND OF THF INVFI~ITION
Field of the Invention The present invention relates generally to the processing of volumetric capilla, y cytometry data for the purpose of counting and characterizing cells or cell constituents in a volume of material.

Desc, ;~tio~ of Related Art The optical analysis of biological specimens, such as blood, has widespread applicalions. There are many computer controlled instruments in the market for providing such analysis, including flow cytometers, automated blood cell analyzers and blood cell classifiers.
As described in the above cross-referenced application, a volumetric cytometry instrument has many advantages. In particular, the amount of blood being analyzed is controllable, the handling of the blood is reduced, and analyzed samples of blood can be stored for further processing.
However, the processing of blood samples in a volumetric system raises a number of problems. Particularly if counting of blood cells is required, it is necess~ry to analyze the entire volume in the capillary or other container which holds the material to be analyzed. Blood cells on the side of a container may be difficult to detect, if the ar!alysis instrument is not calibrated for the precise dimensions of the container.
For instance, in Kamentsky, U.S. Patent No. 5,072,382, samples of blood were a, F l ~ to a slide. A region to be analyzed was defined by sy"ci~ror,ization pulses in the scanr.ing apparalus. (See column 14, lines 49~3 of Kamentsky). Synchronization pulses in the scanning mechanism cannot be precisely aligned with a container such as a capillary or cuvette for a blood sample because of the variations in the shapes of such con~;ners, and va, ia~ions in the alignment of mounts for Anomey Oock-~ No.: B~ H
mah/bmi/2002.002 - 21g8204 the containers in the scanning mechanism. It will be appreciated that the ability to precisely mount and manufacture containers is quite advanced. However, the scanning of containers for the,purposes of processing cells may require resolution on the micron scale.
An additional difficulty arises because of the charac~erislics of dyes used to mark target cells. For instance, when analyzing cells for the presence of specific antibodies, it is common to tag the cells with dyes which fluoresce with a particular spectrum in response to an excitation beam. If more than one antibody is to be detected, more than one dye is used. However, the fluorescence spectrums of various dyes may overlap. Thus, it is difficult to fully process the informatioll in detected fluorescence generated by plural dyes with overlapping spectra.
Furthermore, when it is necessary to count a particular number of target cells within a volume, to achieve a slatistically valid count, a relatively large sample must be used. A large sample of blood, when it is scanned on the micron scale, can generate very large amounts of data. It is important for practical analysis machines that the data be processed in a reasonable amount of time. For instance, as described in the above cross-referenced application, the sample can scan for the presence of two dyes with overlapping spectra, with two,channels of data. Each channel of data includes information relevant to both dyes.
Further, the scan involves about 10,000 lines of 200 pixels each, resulting in 2 million samples per channel, which for 2 b~tes per sample in 2 channels amounts to a total of 8 megabytes of raw data.
Furthermore, the fluorescence monitoring techniques are susceptible to a low signal-to-noise ratio. Thus, it is important to be able to process these large amounts of data with high bacl~ground noise to Atlorn-y Dockot No.: e~ H
mah/bmi/2002.002 214820~

accurately characteri a and identify target cells within the volume, particularly when unbound antil)~y is present.
Accordingly, it is desirable to provide a method a~nd apparatus for processing data from a volumetric cytometry system which is robust and accurate. Further, the system should be relatively fast and operate in a system having a relatively low memory requirement.

SUMMARY OF THF INVFI~ITION
The present invention provides a method and an apparatus for analyzing a sample of cells or cell constituents, including but not limited to blood, within a capillary in a volumetric cytometry system. According to the invention, the sample cells or cell constituents have one or more detec~ characterislics. The system provides for detecting the edges of the container, counting the cells or cell constituents within the container, charac~eri~ing the cells or cell constituents within the container, and evaluating channels of data which contain information relevant to more than one of the detect~hle characteristics of the cells or cell constituents.
Accordingly, the present invention can be characterized as an apparatus which comprises a scanner for scanning a container of sample including target cells or cell constituents. Data sampling circuitry is coupled to the scanner to genera~e scanned images of the sample in the container. According to one aspect, the scanner and data sampling circuitry produce a plurality of channels of data, and a corresponding plurality of scanned images. A processing system is coupled to the sampling circuitry, and includes resources to count and/or charac~eri~e the target cells or cell constituents in response to the scanned images.
These processing resources, according to one aspect, are capable of processing the scanned image, which includes information At~om y Oock~t No.: BM12002MAH
mahhmU2002.002 , relevant to more than one of the dete.,~hlo characleris~ics of the cells or cell constituents, to distinguish such detectable cl~ara.,terislics. Such resources may include software for ~e, ror"~ing a correlation analysis between the scanned image having information relevant to more than one chara~terislic and anoll ,er scanned image in the plurality of scanned images. In one prefer,ed system, two scanned images are generated based on fluorescel~ce data from dyes that have overlapping spectra. The two scanned images are processed using a linear regression analysis among corlesponding pixels in the scanned images near target cells to characterize relative cGntents of two fluorescing dyes in a target cell or cell constituents.
AccGr~ing to another aspect of the present invention, target cells or cell constituents are identified from the scanned images using processing resources which identify a peak pixel within a neighborhood, and compare the amplitude of the peak with the amplitude of pixels on the perimeter of the neigl,borhood. If the peak pixel value exceeds the perimeter pixel values by more than a predetermined threshold, then the resources characterize the neighborhood as containing a target cell.
Upon identifying a target cell in this manner, segments of data from the plurality of scanned images corresponding to the identified cell can be saved for further analysis, such as the linear regression analysis discussed above.
f urther, according to another aspect of the present invention, in addition to determining a relative con~, ibution from more than one dye in a scanned image, a parameter indica~ing the in~ensity of the fluorescence of a target cell or cell constituents is determined by filtering the identified segments of data from the plurality of scanned images based upon the expe~ted cl)aracteristics of target cells or cell constituents. For example, in one novel species of the invention, the Anomey Oocket No.: E'"~'~H
mal ~bmU2002 002 214820~

segmer,ts are fiKered by defining a neiyhbo, hood of pixels for each identified segment in the scanned images wherein the neigl ,bo, hood is larger than the e~p6~ted size of the target cell. The pixels within the neighbGrl,ood are processed to compensate for background noise and generate an intensity value for the target cell within the neighborhood based solely on pixel values within the nei!Jhborl ,ood. For instance this processing may involve a matched filter multiplying the intensity values of the neiyhborhood of a cell by a set of values which reflects the expected i"tensi~y profile of a typical cell. The resulting products are summed to yield an amplitude estimate which optimizes signal to noise ratio for that cell. The perimeter of the neighborl,ood is determined based on the expected shape of the target cells or cell constituents within the nei~ orhood.
According to yet another aspect of the present invention the processing resources pei~o"n edge detection and ignore contributions to the scanned images which fall outside of the deteàed edges of the container.
In the prefel,e~ system the cells or cell constituents are characterized using a slope value deterrrlined from the linear regression analysis over a neiyhbo, i ,ood defined for a target cell between two scanned images of the cell generated from overlapping fluorescence spectra of two dyes. Using the linear regression analysis a "slope value" is determined for each target cell. This slope value is then multiplied by the intensity value of the neighborhood in one of the scanned images to produce an analysis coordinate. The cell or cell constituents is characteri~ed based upon the position of the analysis coordinate on a chara~:teri~dlion graph. The characte, i~alion graph is defined with a first region within which target cells or cell constituents having one dye should fall a second region within which target cells or Attorney Dockd No.~ H
mah~i~7 n~

21~8204 cell constituents having the second dye should fall, and a third region within which target cells or cell constituents stained with both dyes s~ould fall. The regions are defi- ,ed based on the background signal cl ,aract6ris~ics of the scanned images for the purposes of signal immunity and more accurate characteri~ations.
The present invention can also be characl~ri~6d as a method for analyzing the sample within such a container. The method includes the following:
scanning the material with a detector to generate a plurality of channels of data, in which at least one of the channels may contain inforrrlation relevant to more than one of a plurality of detectable characteristics of the target cells or cell constituents;
sampling the plurality of channels of data to produce a plurality of scanned images of the sample; and 15 analyzing the plurality of scanned images to characterize the target cells or cell constituents in response to the plurality of channels of data, including processing the scanned image corresponding to the one channel which includes information relevant to more than one characteristic to distinguish such detectable characleristics, processing at least one of the plurality of scanned images to identify segments in the plurality of scanned images con~aWng target cells oç cell constituents, filtering identified segments of data based upon expected cllar~c~erislies of target cells or cell constituents to generate respective intensity values for the iden~iried segments, and characterizing the target cells or cell constituents based on the intensity value in at least one of the plurality of scanned images for a particular segment, and a value based on correlation analysis (such as the slope in a linear regression analysis) between two scanned images of a segment of data.

Attom~y Dodt-t No.: Q'"~'~H
,.~h~ ~ ~

214820~ 1 `._ The system may also mclude analyzing at least one of the scanned images to detect the edges of the container and ignoring data found outside of the det~ d edges. Furthermore, th~ process of characleri~ing the in~e"sit~ value for a particular segment of data may include defining a nei!JhLo, hood of pixels for each identified segment the neighborhood being larger than the ~Ypected size of the target cell and processing the pixels within the neighbo, hood to compensate for background signal and generale an intensit~ value for the target cell or cell constituents within the neighborhood.
Other aspects and advantages of the present invention can be seen upon review of the figures, the detailed desc, i,l~tion and the claims which follow.

BRIFF IlFSCRlPTlON OF THF FIGUF?FS
Fig. 1 is a schematic block d;agra,., of a scanner apparatus with a data processing system according to the present invention.
Fig. 2 is a schematic block diagra,., of the data processing system used in combinalion with the system of Fig. 1.
Fig. 3 is a schematic diayrar~, illuslrating the scanning process used in the scanner of Fig. 1.
Fig. 4 schematically illustrates an organization of data in the scanned images generated by sampling the output of the scanner of Fig. 1.
Fig. 5 is a plot of a representative scanned image from the system according to Fig. 1.
Figs. 6A and 6B together make up a flow chart for the basic data processing loop for the system according to the p,esent invention.
Fig. 7 illust. ates the overlapping spectra of two dyes which are analyzed according to the present invention.

Attome~ odcd No.: F'~
mah/bmUX~002 -Fig. 8 is a graph illust~aling the linear regression analysis used in the characte, i alion of cells according to the pres~nt invention.
Fig. 9 illustrates a cell classilioalio" graph used for classifying cells based on chann~ls of data which include ove, lapFing information, S accordi"g to the presenl invention.
Fig. 10 is a flow chart illuslfaling the generation of background noise indices according ta the p~5~3nt invention.
Fig. 11 is a flow chart illuslr~ling the edge detection process according to the present invention.
Fig. 12 is a flow chart of the process used for detecting cells according to the present invention.
Fig. 1~ IS a flow chart illustrating the ~.rocess for linear regression analysis accord;,)g to the present invention.

DFTAII Fn DESCRIPTION
A detailed description of prefer,ed embodiments of the present invention is provided with respec~ to the figures. Figs. 1 and 2 illustrate a hardware environment for the present invention. Figs. 3-13 provide an explanation of the processing resources used to count and characterize target cells or cell constituents according to the present invention.
As can be seen in Fig.1, an apparatus for volumetric capillary cytometry is provided. The machine is designed to process material within a capilla~y 10, which has a known volume. In the prefer,ed system, the capillary may be rectangular in cross-section, having a width of about 0.4 millimeters to 1.5 millimeters, a length of about 40 millimeters, and a depth of about 25 to 225 microns, and in one embodiment, 100 microns. This capillary is suitable for detecting or characterizing a variety of cells or cell constituents. In one embodiment.

Attomey Dockd No.: BM1200~MH
" ,dl Ll ~ ~

219820~

.._ it is used for the eharacte~ i~ation of a CD3/CD4 assay, in which concent~tion of CD3 and CD4 anti~edies are to be determined. In this type of assay, there are typically three populations present. CD3 positive cells stained with a CyS l~hs"~ antibody only; CD4 cells S stained with both a Cy5 l~h~ d CD3 antibody and a CyFr labelled CD4 anli~Jo-~y. Monocytes are stained with the CyFr labelled CD4 antibody only. It is found that the cells of int~resl in this assay are about 10 microns in diameter and are well classified using a capillary of the dimensions outlined above. It will be appreciated that the present invention is not limited to the c~,araderi~dlion of a CD3/CD4 assay.
According to one method useful with the invention, a biological fluid, such as~whole unco~ lated blood, can be reacted with an excess amount of a binding agent that contains a fluorophone excitable at a given wavelength. The fluorescently-l~helsd binding agent is selected to react with binding sites present within the sample. For example, a fluorescently-labeled antibody directed to the CD4 cell surface marker present on some leukocyte blood subcl~sses is reacted with a sample of whole blood. The lah~le~ binding agents and the binding sites, i.e.
Iabeled anti-CD4 antibodies and the surfaces of CD4-bearing leukocytes in the example, form fluorescent complexes that will emit a signal when used with the apparatus of the present invention.
After the fluidic sample is reacted with the labeled binding agent, it is diluted and the into capillary 10. Minimal processing of components of the biological fluid nor separation of bound and unbound binding agent is required at any point in the praclice of the method of the present invention. An optical scan is made of the sample in a volumetric manner and fluorescence emission is sequentially recorded from each illuminated columnar region.

AKomey Docket No.~ H
mah/bmi/2002.002 21~8204 Fluorescence emission occurs from both the binding agent-binding site complexes and from the free binding agent but a more if~tense signal relative to background level comes from areas where the binding agent is clustered, i.e. cells or cell constituents exhibiting binding sites to which the binding agent is d;rected. Therefore, a signal of heightened fluorescence cGr,aspGnds to a cell or cell constituents, and is recorded as such. When the fluorophores with which the anti-CD4 antibodies are lAheled and excited in the given example, the fluorescence emitted and recorded as an event sign;~ies the presence of a leukocyte that expresses the CD4 antigen.
The enumeration may occur in an absolute volume, depending on a desired application, by noting the beginning and ending points of the lengthwise scan of the capillary tube and measuring incremental steps therebetween. This quanUtalion of all of the fluorescent targets in a fixed, precise volume is a powerful method of quickly obtaining detailed population data.
Fluorophores that activate at dilrerenl wavele.-~ths can be combined with binding agents directed to dilrerent binding sites, so that the presence of multiple reaction moieties in the sample can be detected. From the precise known volume of the capillary tube that has been scanned, a quick reading will identify the number o~f cells or cell constituents of a particular subclass per unit volume that are present in the sample. To illustrate, this method can quickly distinguish and enumerate the monocyte granulocyte or Iymphocyte subsets of a given volume of a blood sample through reaction of the whole sample with differentially-excitable fluorescently-labeled antibodies d.recled to the cell surface an~igens. The T-cell leukocytes can be directed to CD3, CD4, and CD8. The optical system is simply set to excite each fluorophore at its crucial wavelength and a detection channel is created Attorney Dock~t No.: 8Mi200~AH
mahlbmU2002.002 2I4820~
_ to correspond to the emission wavelen~ll, of each fluoropl,Gre.
Al~e")dlively, a ratio can be obtained without counting a precise volume, e.g. this is a rapid technique for obtaining CD4/CD8 T-c.ell ratios, important in determining the progression of AIDS.
S When an assay is l~e,ror",ed to determine leukocyte subclasses in whole unco~gu'ated blood using the lechnique of the present invention, a two or three minute wait between placement of the reacted sample into the capillary tube and the optical scan allows for the natural density of the numerous red blood cells present in the sample to cause settling of the red blood cells to the bottom of the capillary tube and the sl ~hse~uent displacement of the white blood cells. This natural buoyant effect causes a resultant localion of the white blood cells near the upper portion of the capi'l2~ tube and assists in fluorescence detection l~ecAIJse of the top-down scan geometry of the presenl invention.
Because of this effect, coincidence of targets is also negligible.
The scanner is based on use of a laser 11, such as a helium -neon laser (as shown), an ion laser, a semiconductor laser, or the like, which generates a laser beam along path 12. Laser 11 preferably emits in the 600 to 1000 nm range. The laser beam along path 12 passes through a beam splitter 13, such that a portion of the beam is diverted to a power meter 14 for monitoring the laser output power., The ;nain beam passes through the beam splitter 13 to a narrow line filter 15, which selects the wavelength of interest generated by the laser 11.
Next, a dichr~ic beam splitter 16 receives the beam which passes through the filter 15. The dichr~.~ beam splitter diverts the selected output from the laser toward a steering mirror 17. The steering mirror 17 diverts the beam to a folding mirror or prism 18. From the folding mirror or prism 18, the beam is d;re~ed to a scanner 19.

Attorney Docket No.: e'"~'~H
, n..h/~ ~ ~

2I 1820~

The scanner 19 allows transverse and longitudinal scanning of the laser beam aaoss the sample capillary 10. The scanner assembly includes a galvanometer mounted mirror 25 which rotat~es a few degrees back and forth in a rapid fashio n at about 20-200 Hz (peak-to-peak 6 12). The beam is defleded by the galvanometer mounted mirror 25 to a first lens 22, through a second lens 23 and through an objective lens 24 to the capillary 10. Two lenses 22 and 23 are desig"~d to be confocal, that is, they are sepa~at~d along the optical path by their focal length, and they have equal focal lengths. It is not necessAry that the lenses be confocal, but they must have ove, lap,~i.)g focal planes.
Similarly, the distance between lens 23 and the objective lens 25 of the microscope rnust be precisely conlfolled so that the beam, as it is rotated by the galvanometer mounted mirror 25, appears to be rotating about a virtual point directly in front of the micr~scope objective.
As schematically illustrated by ghosted outline 39, the scanning assembly 19 is designed to move longitudinaily along the capillary 10 for a distance of about 40 millimeters.
A variety of other scar-ner mechanisms can be used as suits a particular application of the invention.
The whole scanner assembly 19 is con~ by computer 30, as schematically illustrated by line 26.
The beam impinging upon the outer wall of capillary 10 traverses the wall and illuminates a columnar region of the sample causing fluorescent emission from the sample. Light collection occurs in an epi-illumination manner. The emitted fluorescence is collected by microscop~ objective 24 and ~i,ec~ed back, as a retrobeam.
Microscope objective 24 has a central portion for p~ssage of incident beam and uniform depth of focus of the inc;dent beam through capillary Atlorney Qockot No.: 3M1200~UAH
malVbmU2002.0~2 - 214820~

10. BecalJse flolJrescenl emission is over a very wide angle, flourescent ~ol'ec-tion occurs over a wider portion of microscope objective 24.
Fluor~sc6"ce given off by the dyes in re3pollse t~o the exGita~ion beam generated by the laser 11 retl aces the optical patn through the scanning mechanism 19, the prism 18, the steering mirror 17, to the dichroic beam splitter 16. The fluorescence comes from the focal point of the microscope objective 24 and is collimated with a diameter of about 8 millimeters as it comes out of the microscope objective. The d;ch,a9c beam splitter 16 allows the fluorescing wavelengths to proceed along path 26.
The beam on path 26 enters a bandpass filter 27, or a series of the same, whlch are designed to filter the backcc~lering of the laser beam itself which are due to weak reflections from the optical elements and surfaces of the capillary. These r~le~lions may be much stronger than the actual fluorescence that is detecled from the sample.
From the bandpass filters 27, a folding mirror 28 directs the beam through a focusing lens 29. The focusing lens brings the colimated light from the sample into a focus, and through a pinhole filter 30. The light from the capillary, which arises outside the focal point of the microscope objective 25 will not be columnated when it enters the lens 29. Thus, the pinhole filter 30 rejects fluorescence from regions that are not of interest. The pinhole size is chosen so as to define a volume from which to collect fluo, escence intensity from the sample. Typically, this volume is ss'~te~ to be about five or ten times the expected volume of the target cells.
Through the pinhole 30, the beam enters a pl ,olo" ,ultiplier box 31. The photomultiplier box includes a dichroic beam splitter 32 which separates the detected fluorescence into two basic components. The first component, channel 0 is light having wavelengths below about 680 Attom-y Dockd No.: B~"m~'~H
mahlbmU2002 002 21~8204 _ nanometers in the pr~sent embodiment. The second component, channel 1, is light having wavelengths above about 680 nanometers.
Cl,snnel 0 is directed to a first ~hotG."ultiplier 33. Channel 1 is directed to a second photomultiplier 34. rl ,otomulli~ rs are connected across line 35 to the computer 40.
Also incl~ed in the mechanism, but not shown, is an autofocus mechanism. The autofocusing system uses an algor~thn, which involves measuring the fluorescence in a preliminary scan. At a first position in the capillary, the microscope objective focus is scanned to find the position of maximum fluorescence. This value is stored, and the objective is moved to a second position in the capillary. Again, focus in this position is scanned for maximum fluorescence. That value is stored. Using the two values, as a beginning and end point for the scan, the microscope focus is linearly exl,apolated between the two to optimize the fluorescence reading along the length. In addition, the length of each scan line is set to be slightly larger than the width of the interior of the capillary. This is done to insure that every cell is detected by overscanning capillary dimensions, and later detecting the edges of the capillary to filter irrelevant information.
Fig. 2 schematically illustrates processing resources in the computer 40. The computer 40 includes a CPU 41 cou~led through a system bus 42, as schematically ill-,sl,aled. On the system bus 42 are a keyboard 43, a disk drive 44, or other non-volatile memory system, a display 45, and other peripherals 46, as known in the art. Also coupled to the bus 42 are a program memory 47 and a data memory 48. The output of the ~I,olo",ultipliers, channel 0 and channel one, are supplied on lines 35~ and 35-1, respectively, through analog to digital converters 49-0 and 49-1. The outputs of the analog to digital converters are 16 bit pixel values of the analog signals from channels 0 and 1. These values At~om-y Oockct No.: ?''~'t.H
n~ t 7nno r~

are supplied through a direct memory access (DMA) circuit 50 which ~r~nsrers the pixel values into the data memory 48.
The processing system 40 includes resources that store scanned images of the data for channel 0 and 1, buffers used during the processing of the data, and memory for storing cell data once the cells or cell constituents are Iscate~l cl ,aracteri~ed and/or classified.
Similarly, the progra", memoly 47 includes resources for detecting the edges of the capillary, counting and locdling target cells or cell constituents within the scanned images, cl~aracleri~ing the target cells or cell constituents, and repo, (ing results. More details concerning the processing resources in the computer 40 are provided below with respect to the flow charts and graphs of Figs. 3-13. As will be appreciated, processing resources may be implemented with hardware, software, or a combination of both, as suits a particular use of the 1 5 invention.
Fig. 3 illustrates the scanning technique used for gathering data from the capillary. The capillary 10, as illu .ll.ated in Fig. 3, has a width of about 0.667 millimeters, and a length of about 40 millimeters. A
galvanometer scanning system scans the laser beam along a track 1 on a line longer than the capillary 10 is wide. At the end of line 1, the beam snaps back to the begin"ing of line 2 and scans line 2. The distance between the center of lines 1 and 2 is about 4 microns in the present embodiment.
For a capillary of about 0.667 millimeters in width, the scan lines are about 0.8 millimeters long. This provides 200 four micron samples along each scan line. For a 40 millimeter long capillary, with scan lines separated by 4 microns, about 10,000 scan lines are collected for each blood sample.

Attorn-y Docl~-t No.~ H
mah/bmil20~..002 21~820~

The analog to digital converters sample the scanned data at a rate which creates a pixel value representing fluor~scence in a spot having dimensioos of about 4 microns by 4 microns. Fig. 4 illus~rales a 7X7 neighborl,ood of pixels. Thus, in the upper left hand corner, pixel in row 1, column 1, is found. In the upper right hand comer, pixel in row 1, column 7, is found. In the center of the n~ llbo-i,ood of pixels, pixel row 4, column 4, is found. Similarly, in the lower right hand corner, pixel in row 7, column 7, is found. Fig. 4 also illustrates the size of the laser spot relative to the sample dimensions. In the prefer, ed embodiment, the laser spot (e.g., 50) has a diameter of about 10 microns. Thus, oversampling occurs. That is, the laser spot 50 excites a region of 10 microns in diameter for the pixel at row 7, column 1. At row 6, column 1, a second spot 51 illuminates a region 10 microns in diameter which substantially overlaps with the spot 50 for row 7, column 1. Similarly, the spot 52 for row 7, column 2, substan~ia:ly overlaps with the spot 50 and the spot 51 in column 1.
Fig. 5 illustrates a portion of a scanned image generated with one of the channels of the present invention. The graph of Fig. 5 is a baseline sub~racted represenldtion, where the baseline is illustrated at line 60. The baseline is essentially the average height of all the lines in the scan region. With this value subtracted, a number of peaks, e.g., peak 61, can be seen in the scanned image. These peaks typically correspond to target cells and are processed as described below. Also, each scan includes a region, generally 62, and a region, generally 63, which lie out~ide the capillary. The baseline 60 can be used to define the edges of the capillary becsu-se of the rapid falloff at 64 and 65 co" esponding with the edges of the capillary. The processing resources characterizing the cells ignore pixel values outside the detected edges.

AnorTley Dockot No.: BM1200~UAH
m-h/bmi/~02.002 214820~
.~
As mentioned above, there are two c~annels detected according to the pr~s~al invention. Fig. 5 illusb ales a single cnannel. There will be a cGr, espGn~ing scanned image from the second channel having a similar profile, however, the amplitudes of the peaks wili dffler depending on the magnitude of the fluorescence delec~6d in the second channel. Also, some peaks may be found in one image but not the other.
The basic data processil-g steps ex~ted by the processing resources are illusll aled in Figs. 6A and 6B. The first step in Fig. 6A is to receive the data from channel O and ~ ,anl ,el 1 (block 100). As the data is received, the DMA circuitry loads it into a buffer in the data memory (block 101). The buffer may be a circular buffer or other data structure used to keep track of the amount of data being received. The algorill"n then determines whether a block of data having a prespecified size has been received (block 102). For the purposes of the Rresent example, about 100 to 150 scan lines may comprise a suitable block size. If a block has not yet been completely receivcd, then the algorithm determines whether the last block from a blood sample has been received (block 103). If it has, then the alg~ritl,tl, is done (block 104). If it has not, the algoriUI", loops into block 101 to continue loading data in the buffer.
When it is detecled that a complete block has been received at block 102, then the algo,itl"" parses the data into a plurality of scanned images by dividing the data into a raster image file ImO for the first channel, and a raster image file Im1 for the second chanl~el (block 105).
Data pre-processing involves reading in and processing the data in blocks. The signal processing requires signed numbers, thus, the unsigned 16 bit data is converted into 15 bit signed data.

Altorn-y Dock-t No.~ H
mah/bmU2002.002 ` _ 2148204 Next the two scanned images ImO and Im1 are summed or averagecJ to g~erale a composite image Im2 and the composite irnage Im2 is stored (block 106).
Next an edge dete~tion algorithm such as described below with respect to Fig. 11 is executed (block 107). The edge deteclion algo,itl"" may be supplemented with an algo,ilh,n for evaluating the results to ensure that no false edges such as might be detected by a bubble in a capillary are found.
The edge detection is done using a baseline profile where the baseline is the average of all the scans in the block ignoring the scans that have peaks higher than a certain threshold After btock 107 the algo,iU"~, p,oceeds to determine thresholds to be used for pa, tic e detection in the composite image Im2. These thresholds may be prespec;fied empirically determined values or they may be adaptively computed for each buffer or each sample. One algo,ill"" for determining the threshold may involve determining the maximum and minimum values for each 7X7 pixel neighbo,l,ood in the block being processed The maximum minus minimum value with the highest frequency of occurrence is used to estimate the threshold for particle deteclion. Distribution of the maximum and minimum for peaks of neighborhoods in the buffer are determined and the thresholds are set so that peaks are detected if the maximum and minimum values in the nei!Jhbo, h ood differ by amounts smaller than a threshold 3 standard deviations below the average peak height.
After determining the lhreshclds for particle detection in block 108 the algorithn, proceeds to compute the background indices for the scanned images ImO and Im1 using an algorit~",~ such as described with respect to Fig. 10 (block 109). Afler computing the background indices the algori~l"" may then proceed to do a baseline su6sl,action step for AttomeyDock-tNo.: EIM1200~AH
~ 7~ ~?

-the scanned images ImO and Im1. This is an optional step, depending on the tech,)iques used for particle deteetiGn and cell cl~ar~cteri~dlion set out below.
One baseline removal technique involves finding the minimum, or average of the N minimum lowest points along a given scan (N equal about 10). The above minimum, or average minimum, is then subt(acted from all pixels in the block, and negative values are clipped to 0. Baseline removal is optional. In particular, if cell detection uses the peak slope criteria, as described below, it is not necess~ry to do baseline removal. However, it may be desirable to have baseline subtraction to eliminate edge effects in an overscanning situation.
After baseline sublraclion for images ImO and Im1, the algorithm does a p81licle detection routine using image Im2 (block 1 1 1). The particle detection process is illustrated below with respect to Fig. 12.
The next block determines whether a particle is detected (block 112). If a particle is detected, then the neigllbGrl,oods of pixels from images ImO, Im1 and optionally Im2 are saved, and cell parameters are computed using data in the nei~l,bG,l,ood of the pixel maps, e.g. in respGnse to the saved neighbo,l,oods. If a particle is not detected, then the algo,itl"" determines whether the Im2 buffer has been completely processed for particle detection (block 1 14). If not, the algorithm loops to block 111 to continue the particle detection routine. If the buffer has been finished, then the algorithm loops to block 102 in Fig. 6A, as illus~a~ed, to begin processing a next block of data.
Thus, for example, the raw data consists of two 200xN (where is less than or equal to 10,000) raster image files for each channel O and channel 1, where each pixel is a 16 bit unsiy, 16d integer, generaled by the output of the analog to digital converters. The image data may be processed in blocks. Data for scan lines at the boundaries of the image At~om-y Dock-t No.: '`'"~'~H

21~8204 -block may be buffered to deal with cells crossing the image block boundaries.
The data block size is 200 pixeis high by N scans wide, where N
is about 128. The buffered block size may be 200 X 16 pixels on the S block bound~. ies. 8y image processing in blocks, a smaller arnount of memory resources are used for the image processing.
The technique of saving the nei,Jhbo,l,ood values from the scanned images for each block, as p~, licles are detected, and then continuing to process ~itional blocks allows real time gathering of data and cell or cell constituent deledion, with ability to compute cell parameters and characte- i~a the cell later, or with a time shared processing technique. This greatly enhances the efficiency of use of the computation resources to allow sampling of very large amounts of scanned data in substantially real time.
Cell and cell constKuent cl~arac~eri~dtion, according to a preferred embodiment of the presen~ invention, involves utilizing information from the two cl ,an,lals. The two ~;hannels in the present system include data which is relevant to both of the dyes which are to be detected. Thus, a technique must be used to discriminate information from the two dyes in the two channels that is efficient and noise immune. Accordingly, one species of the presenl invention applies a linear ~gfessiGn technique over the neighbo,hoods of pixels saved for detec~ pa, t: ~' s.
The problem to be solved can be appreciated with respect to Fig.
7, which schematically illustrates the spectfa of fluorescence for the two dyes detected by cha"nel 0 and channel 1, respectively. Thus, a cell containing an~igens stained by the first dye will fluoresce with a spectrum suc'n as spectrum 120. Similarly, the cell with a dye attached to an a.ltigen of the second type will fluoresce with the spectrum 121.

Anomey Oocket No.: E~
mah/bmi/20~2.002 21~820-~

As can be seen, the two spectra sul,slanli~lly overla,o. The dichroic beam splitter 32 in the phot~multiplier mechanism 31, as shown in Fig.
~, splits the fluore-~cen~ beam along the 680 nanometer line 122 to generate two signals. This line has been empirically determined for the presently described system to provide good separation. Thus, the two signals both contain information which is generaled in response to both dyes.
The linear regression technique utilized to discriminate the information in the two channels is described with reference to Fig. 8. In particular, a 7X7 neighborhood of pixels is saved from each of image ImO and Im1, cenlered on each target cell which is detected using the algorithm of F~lgs. 6A and 6B. The 49 coordinates may be plotted as shown in Fig. 8, where each dot is positioned with the magnitude of the first channel on axis X and the magnitude of the second channel along 15 axis Y, such that sample at row 1, column 1, and sample at row 7, column 7, may appear at the points (1, 1 ) and (7, 7). Similarly, the sample at row 4, column 4, may appear at the point (4, 4). As can be seen, the sample from row 4, column 4, will have the highest average of amplitude from both channels because of the cell classification technique. The dots are then a, pl'8d to a linear regression algorithm, as well known in the art, to find the best fit line 123 to the dot plot, Pres it al., Numerir~l RP~.U; Pe in C. 1988, p. 523. This produces the slope "m" and the offset "a" for each target cell. Thus, the magnitude of the cont, ibution from channel O can be expressed as the slope "m" times the magnitude of the contribution of channel 1, plus the offset value "a".
Fig. 9 is a graph which illusl~ales the technique used for characterizing the detec~ed cells. In particular, the intensity value for a detected cell is de~ined as [CH1 ] for the scanned image Im1 and [ChO]
for the sc~nned image ImO in the respecti~/e 7X7 nei~JhbGrhoods. This Attomey Docke~ No.~ H
mah/bmU2002.002 21~820~
.
intensily value is determined by a matched filtering technique based on the expected cl lafac~l isli~s of target cells over the 7X7 nei~l ;bo, hoods.
A matched filter can bQ constructed from e~pected or typical pixel values in the neighborl-~od of a cell. The matched filter is a matrixl of coefficient which will be multiplied with the conesponding pixels from the neighborhood of the Jete~led cell. The 49 resulting products can be summed to yield a single intensity value (~CH0] or [CH1]) for the detected cell. The cG~ff~-~nts can be chosen to sum to zero, in which case the conslant background signal is cancelled out.
The inlensity value for the cell is then plotted on the graph of Fig.
9 by multiplying the slope determined using the linear regression analysis above by the inlensity value of one of the channels.
As can be seen, the graph of Fig. 9 is divided into five regions.
The first region, 130, is for target cells which are dyed substantially only with the first dye which is centered on channel 1. The second region, 131, is defined for target cells which are dyed substantially only with the second dye. The third region 132 is defined for cells which are believed to be stained with both dyes. The fourth region 133 and the fifth region 135 are a "no call" region to ensure that bad data is ignored.
The first region 130 is defined during calil,ralion of the device by doing a scan of a sample dyed only with the first dye. Linear regression analysis is applied to create a line 136 based on the one dye detection from channel 1. A similar technique is used to define line 137 for the second dye which has most of its spectrum det6.,~ed by channel 0. A
background index is c~l~J~ated for the buffer in question to define regions indicated by dotted lines 138 and 139 below which for channel 1 and above which for channel 0, samples are charac~ei i ed as having only one dye. The samples which fall in the region 132 are characterized as having both. Samples which fall in the regions 133 AnomeyDocketNo.: 9~'12~'~H
rn~h/bmu2w2~oo2 21~820~

and 135, or which have a magnitude value which is too low, are ~;har~eri~e~ as no calls.
Ap,vendi~ A provides a source code for the cell c~!dr~ctefi2ation routine according to one embodiment of the pr~sent invention as a means of providing an example of pfocassing resources which might be used to accomplish this sort of classification.
The goal of cell or cell constituent classifi~3~io,) is to determine the decisiQn boundaries based on pop!Jl~tion~ dependent parameters.
If necessary, the boundaries determined can be validated with population ~dtislics.
In the CD3/CD4 assay, there are 3 cell popul~tions present. The CD3 positive cells are stained with the Cy5 l~hel~d antibody only. The CD4 cells are stained with both the CyS lal-ele~ CD3 antibody and the CyFr l~hele~ CD4 antibody. The monocytes are stained with the CyFr labeled CD4 antil~o~y only.
Since the CD3 positive cells and the monocytes are stained with one dye only, their slope distributions should cluster around the Cy5 and CyFr slopes respectively. The Cy5 and CyFr slopes are determined in the compensation matrix calibralion. The spread or distribution of the clusters can be determined from the background noise estimates.
Similarly, in the CD3/CD8 assay, there are 3 cell populations present. The CD3 positive cells are stained with the Cy5 labeled anli~o~Jy only. The CD8 cells are stained with both the Cy5 labeled CD3 antiL~ody and the CyFr labeled CD8 antibody. The NK cells are stained with the CyFr l~heled CD8 antibody only.
Since the CD3 positive cells and the NK cells are stained with one dye only, their slope distributions should cluster around the Cy5 and CyFr slopes respectively. The Cy5 and CyFr slopes are determined in Attomey Dock-t No.: B"'~'~H
1 Idl Ll : ~}2.002 21~820~
` .

the compensation matrix calibration. The spread or distribution of the clusters can be determined from the background noise estimates.
The following classi~ioa~ion rules may be applie~
Non-Cells S A particle is classified as non-cell, if any of the following criteria is met:
1. Particles with cc ~rela~iGn coefficient (for the regression fit for planes O and 1 ) less than the threshold value (0.8).

2. Channel O value less than background noise threshold.

3. Channel 1 value less than background noise threshold.

Monocytes and NK Cells 1. Cells with channel 1 value grealer than the CyFr slope line minus a constant background noise offset for channel 1.
. 15 CD3 Cells 1. Cells with channel 1 value less than the CyS slope line plus a constant background noise offset for d ,annel 1.

C D 4/8 C ells Any cell that does not satisfy the above criteria are potentially a C D 4/8 cell. Cells that lie too close to the Cy5 or CyFr slopes are labeled as "no-calls". The slopes di~rence between the CyFr and CyS slope are divided into slope regions. Cells that are below the 10% or above the 90% region boundaries are classified as no~alls.
The use of the slope value from the linear regression a ,nalysis provides a noise immunity, and improves the robustness of the system.
This analysis may be replaced by solving two equ~tions with two Atton~y ~OCk-t No.: F~ H
mah~bmi~2002.002 -unknowns based on the intensity values for channel O and channel 1, respectively.
The cross co" ~,lation COerriC;en~ betweG,- col I dsQond;n9 pixels from image O and image 1 is determined. H the signal is dominated S random noise, one would expect a poor correlation coefric;enl. A cell with good signal gives a co"~lation that ranges from 0.9 to 1Ø
The cross CGI I elatiGn technique can be e~ten~ed to correlate individual cêlls with thQ average cell profile. A con,posite average cell profile can be generated by averaging the cell profile of all the cells detecte-J. A good cor, ela~iGn coefficient indicates that the cell shape is similar to the average cell shape. An artifact peak usually has a ~lirreren~ cell shape profile and thus gives a poorer cross correlation coefficient.
Fig. 10 illusllatês an algoriU,I,~ for computation of the background noise indices, used as described above, with respect to Fig. 9, to define the regions around the line for the two dyes within which a cell will be characterized as containing only that dye.
This background index can be computed for each buffer, or it can be computed across the entire image.
The technique involves defining a plurality of bins for N
background baseline noise levels, where N is about 2,000, and each bin 16 bits wide, in one embodiment (block 200). for each buffer, a V\IXW
pixel map surrounding each pixel is derined, such that WXW is large enough to contain almost all the signal from the exrected size of the cell. This may be about 5X5 in the present example (block 201).
For each V\~N pixel map, the maximum and minimum pixel values are determined for each pixel in images Im1 and ImO (block 202).
Each pixel is then assigned an integer bin number based on the dirrerence between the maximum and minimum values within the pixel AKorr~y Dock t No.: BM1200~AH
rn~hhm~O02 21~820~
-map divided by the size of the allocated bins (block 203). As mentioned above, there may be 2,000 bins, 16 bits wide each for a range of about 32,000 values.
Wlth the integer bin number, the count for the co,.~,sponding bin is incremented for that pixel (block 204). After incrementing the bin number, the algori~l)." determines whetl,er there are more pixels to process for the buffer (block 205). If there are, it loops back to block 201. If not, the algGl ith m determines the bin number with the highest count (block 206). The background noise index for this buffer is set to the bin number with the highest occurrence times the bin size (block 207). These indices are then stored for both channels (block 208). As mentioned above, the process is carried out for both channels to achieve two separate background noise indices.
Fig. 11 illustrates an algoritl "" for performing edge detection, as menlio"ed above. Rec~use the blood sample in this example will contain free dye labelled antibody, a distinct background signal results.
This signal is used to locate the boundary of the lumen of the capillary.
The average image Im2 from chal)nel 0 and channel 1 is gene~led (block 300). A specified number N of scans are averaged pixel by pixel to produce an average scan (block 301). The ma~timum and minimum values are determined from the average scan (block 302). Next, a threshold is set based on the difrerence between the maximum and the minimum times a factor which ranges from 0.1 to 0.8 (block 303). This est~l shes a ll,reshald which is a percenlage of the average amplitude of fluorescence from a given scan lins. The scan lines are then searched from center of the scan to the left until the baseline value is less than the threshold for left edge delec~ion. The left edge is then set as the lo~a~io~ where the threshcl~ is crossed (block 304). Similarly, the algorith,n searches from the center of the scan to the right until the Altomoy Dockot No.: E~ H
m~h~bmV2002 002 21~820~

threshold for right edge detec~ion is pAsse~l The lo~a~ion of the right edge is then defined as the localion where the threshold is crossed (block 305). Using edge detec~ion, sc~nned pixels out.s,ide of the detec~ed edges are ignored in the additional processing des~ ibed above.
Fig. 12 illusbdt~s the algo.iU,.~, for detec(ing a cell or cell constituent. The cell dele~ion algorithm uses the image Im2, which is based on the average (or sum which for this purpose is substantially the same thing) of images ImO and Im1 (block 400). A 5X5 pixel map surrounding each pixel is then detined in sequence (block 401). This pixel map is tested. Ihe test is a five step test whidl involves determining whether the center pixel is the pixel having the highest value of all pixels in the pixel map. If it is, then the test takes the difrarence bet Necn the center pixel and the top left comer pixel and determines whether this di~ference is greater than a threshold which is assigned for cell or cell constituent detection. Next, the center pixel and the top right comer are used to make the ll,resho'i determination. Next, the center pixel and the bottom left corner pixel are used to make the threshold determination. Next, the center pixel and the bottom right corner pixel are used to maker the tl,resho'd determination. If the center pixel has the highest value, and anotl,er pixel in the map has an equal value, then the algorithm passes the test. In order to avoid counting a cell twice, in this inslance, the center pixel value is incremented by one, so that when the other pixel value having the high value is encountered, It will be determined that it is not the highest value pixel (block 402).
If all five conditions have been satistied (block 403), the x and y coordinate for thé center pixel is saved, and a 7X7 pixel map, or neighbo, hood, surrounding the center pixel is saved from each image AKom-y Dockd No.: B"'~'~l ..: ~ ~qO2 21~820~

ImO and Im1 in the cell parameter list (block 404). These values can be used for later processing of the data as des~ il~ed above.
Using the averaged (or summed) image for cell detection provides better cell resolution, because of magnitude of the signal from any one of the two dyes may be very low on a given cell. This cell or cell constituent det~ion is based on the maximum an minimum values in the neighbG- hood around the detected cell, rather than an absolute peak value. This provides immunity from background levels may vary over the scanned region.
Fig. 13 illustrates the basic algoritl "" for linear regression analysis of the 7X7 pixel map. The linear reg~ssion analysis begins by taking the 7Xrpixel map of a detected cell from both images lO and 11 (block 500). A linear regression line is fitted for corresponding pixel points from images ImO and Im1 over the indices for the row i and the column j, such that Im1 (i,j) approximates to the slope times ImO (i,j) plus an offset (block ~01). After fitting the regression line, the slope, offset and CGI, elation coefficient r2 (goodness of ft) are stored in the cell parameter list (block 502).
Thus, a linear regression line fit is computed between the corresponding pixels from image O and image 1 and the slope and goodness of fit values are obtained. The slope determin,ed indicates whether the cell is stained with 1 of 2 antibGJies. Using a 7X7 pixel map to determine the slope gives a better estimate of the slope of the cell.
The goodl ,ess of fit indicates the gooJI ,ess of the data. The advantages of the linear regression analysis include noise reduction, a good estimate of signal quality, a system insens~ti~/e to baseline subtraction, and the well defined oi U~ogonal coordinates.
In sum, a method and apparat~ls is provided for processing data generated by one or more ;hannels of data, where the channels include Anom y Dock-t No.: g~ H
mahhni~.O02 information relevant to more than one characlerislic to be determined, and are taken from a container. The technique allows for counting and ct)atact~ ing the cells or cell constituents within the contained region with minimum ope,alor handling of the samples, repe~t~l~i'ity, and efficient utili~dlion of processing resources.
The data processing resources accomplish data collection, image averaging for capillary edge and cell detection opera~i~ns, background noise determination, baseline removal, and cell charac~e, i~ation.
The data from a detected cell or cell constituent is then extracted by saving it into a cell parameter list in memory. This allows continuous scans of large volumes of data, with processing of thè data saved for later steps when more processi"g resources may be available. Also, it allows reanalysis of data for detected cells in the future based on much reduced file sizes, as compared to what would result if the entire 200 by 10,000 pixel file must be saved for later analysis.
The present invention provides a system for processing scanned data which is very robust and accurate. It allows concurrent data ccl'~ction and analysis, as well as analysis of data after collection.
Further, it allows for completion of data analysis very rapidly, shortly after completion of the data collection systems.
- The for2~ing desuiplion of a prere"ed embodiment of the invention has been presenled for purposes of illualldliG,i and description. It is not inle,)-led to be exhaustive or to limit the invention to the prec;3e forms disclose~ Obviously, many mod~ficalio"s and variatiol,s will be apparenl to pra~,1ilioners skilled in this art. It is intended that the scope of the invention be defined by the following claims and their equivalents.

Attomey Dock-t No.: ~'"~'Ul . . ,; ~ .:~2.002 21~820~
._ APPFI~IDIX A
Copyright Biometric Imaging, Inc., 1994 SOURCF CODF FOR CFI I CHARACTFRI7~TIQN

const float kCellCo.,~' YcnCoefThresha~ 0.8;
CELL-TYPE CCell::CellC-b ~ni' ~ ~ffc n( ) {

float rnaxSlope - CyFrSlope - ((CyFrSlope-CySSlope) 0.1);
float minSlope ~ CySSlope + ((CyFrSlope-CySSlope) 0.1);
if ( (cellCo m 'a' nCoef ~ kCellCo, . ~ 'aflc nCoefThreshokl) l l (peakValueO ~ noiseTh.~73h~'~0) ll (peakValue1 ~ noisemr~h ~
retum (NON_CELL); ll noise peak else if ( (peakValue1 ~ (CyfrSlope peakValueO ~ noiseLever) ) retum (CYFR_POSlTlVE); //~.~no. ytl~ or NK cells else if ( peakValue1 ~ (Cy5Slope peakValueO ~ n - ~LAvel) ) return(CYS_POSlTlVE); /ICD3 celk else if ( (cellSlope ~ rnaxSlope) l l (cellSlope ~ rninSlope) ) retum(NO CALL); llNocall else retum (CYFR_CYF_POSITIVE); ll CD418 cell ) AKor~ k t No.~ H

Claims (58)

1. An apparatus for analysis of a sample of cells or cell constituents, the cells or cell constituients having detectable characteristics; comprising:
a scanner which scans the sample to generate a channel of data containing information relevant to the detectable characteristics;
sampling circuitry, which samples the channel of data to generate a stream of pixel values for a scanned image of the suspension;
a processing system having memory and coupled to the sampling circuitry, including resources which load the stream of pixel values into a pixel map buffer in the memory, analyze the pixel map buffer to identify certain cells or cell constituients in the scanned image, and save a neighborhood of pixel values for each identified cell or cell constituent for later processing.

2. The apparatus of claim 1, wherein the scanner generates two channels of data, and the sampling circuitry generates two streams of pixel values for respective scanned images of the suspension, and wherein the resources of the processing system save a neighborhood of pixel values from both scanned images for each identified cell or cell constituient for later processing.

3. The apparatus of claim 2, wherein the resources which identify certain cells or cell constituients are responsive to the two streams of pixel data.

4. The apparatus of claim 2, wherein the processing system includes resources for characterizing the identified cells or cell constituients using information in the neighborhoods of pixel values saved from both scanned images.

5. The apparatus of claim 4, wherein the resources for characterizing perform linear regression analysis of corresponding pixel values from the two neighborhoods saved for a particular identified cell or cell constituient.

6. The apparatus of claim 1, wherein the resources which analyze the pixel map buffer to identify certain cells or cell constituients in the scanned image, include resources for detecting edges of the sample and ignoring pixel values outside the detected edges.

7. The apparatus of claim 1, wherein the processing system includes resources for characterizing the identified cells or cell constituients using information in the neighborhood of pixel values saved from the scanned image.

8. The apparatus of claim 7, wherein the resources for characterizing the identified cells or cell constituients include a filter based upon an expected characteristic of the cells or cell constituients, and resources responsive to the filter to generate an intensity value for the identified cell or cell constituent.

9. The apparatus of claim 8, wherein the resources for characterizing the identified cell or cell constituient include resources responsive to the filter to generate an intensity value for the identified cell or cell constituient, independent of other saved pixel values.

10. The apparatus of claim 1, wherein the cells are blood cells

11. The apparatus of claim 10, wherein the cells are labeled with a fluorophore excitable in the 600 to 1000 nm range.

12. An apparatus for analysis of a sample of cells or cell constituients, the cells or cell constituients having detectable characteristics; comprising:
a scanner which scans the sample to generate a channel of data containing information relevant to the detectable characteristics;
sampling circuitry, which samples the channel of data to generate a stream of pixel values for a scanned image of the sample;
a processing system having memory and coupled to the sampling circuitry, including resources which load the stream of pixel values into a pixel map buffer in the memory, analyze the pixel map buffer to identify certain cells or cell constituients in the scanned image, and save a set of cell or cell constituient parameters calculated from the pixel values.

13. The apparatus of claim 12, wherein the scanner generates two channels of data, and the sampling circuitry generates two streams of pixel values for respective scanned images of the sample, and wherein the resources of the processing system save a set of cell or cell constituient values from both scanned images for each identified cell or cell constituient for later processing.

14. The apparatus of claim 12, wherein the resources which identify certain cells or cell constituients are responsive to the two streams of pixel data.

15. The apparatus of claim 12, wherein the processing system includes resources for characterizing the identified cells or cell constituients using information in the sets of parameters of pixel values saved from both scanned images.

16. The apparatus of claim 15, wherein the resources for characterizing perform linear regression analysis of corresponding pixel values from the two sets parameters saved for a particular identified cell or cell constituient.

17. The apparatus of claim 12, wherein the resources which analyze the pixel map buffer to identify certain cells or cell constituients in the scanned image, include resources for detecting edges of the sample and ignoring pixel values outside the detected edges.

18. The apparatus of claim 12, wherein the processing system includes resources for characterizing the identified cells or cell constituients using information in the set of parameters of pixel values saved from the scanned image.

19. The apparatus of claim 18, wherein the resources for characterizing the identified cells or cell constituients include a filter based upon an expected characteristic of the cells or cell constituients, and resources responsive to the filter to generate an intensity value for the identified cell or cell constitutient.

20. The apparatus of claim 19, wherein the resources for characterizing the identified cell or cell constituient include resources responsive to the filter to generate an intensity value for the identified cell or cell constituient, independent of other saved pixel values.

21. The apparatus of claim 12, wherein the cells are blood cells.

22. The apparatus of claim 21, wherein the cells are labeled with a fluorophore excitable in the range of 600 to 1000 nm.

23. An apparatus for analysis of a sample of cells or cell constituients having detectable characteristics; comprising:
a container having detectable edges and a lumen for the sample;
a scanner which scans the container to generate a channel of data containing information relevant to the detectable characteristics of the sample in the lumen of the container;

sampling circuitry, which samples the channel of data to generate a stream of pixel values for a scanned image of the sample;
a processing system having memory and coupled to the sampling circuitry, including resources which load the stream of pixel values into a pixel map buffer in the memory, analyze the pixel map buffer to detect edges of the lumen, and analyze the pixel map buffer to identify certain cells or cell constituients while ignoring pixel values in the scanned image outside the detected edges of the lumen.

24. The apparatus of claim 23, wherein the processing system includes resources to save a neighborhood of pixel values for each identified cell or cell constituient for later processing.

25. The apparatus of claim 23, wherein the scanner generates two channels of data, and the sampling circuitry generates two streams of pixel values for respective scanned images of the material, and wherein the resources of the processing system save neighborhoods of pixel values from both scanned images for each identified cells or cell constituients for later processing.

26. The apparatus of claim 25, wherein the resources which identify certain cells or cell constituients are responsive to the two streams of pixel data.

27. The apparatus of claim 25, wherein the processing system includes resources for characterizing the cells or cell constituients using information in the neighborhoods of pixel values saved from both scanned images.

28. The apparatus of claim 27, wherein the resources for characterizing perform linear regression analysis of corresponding pixel values from a neighborhood for both channels saved for a particular identified cells or cell constituients.

29. The apparatus of claim 23, wherein the processing system includes resources for characterizing the identified cells or cell constituients using information in the neighborhood of pixel values saved from the scanned image.

30. The apparatus of claim 29, wherein the resources for characterizing the identified cell or cell constituient include a filter based upon an expected characteristic of the cells or constituients, and resources responsive to the filter to generate an intensity value for the identified cell or cell constituient.

31. The apparatus of claim 29, wherein the resources for characterizing the identified cell or cell constituient include resources responsive to the filter to generate an intensity value for the identified cell or cell constituient, independent of other saved pixel values.

32. The apparatus of claim 23, wherein the cells are blood cells.

33. The apparatus of claim 32, wherein the cells are labeled with a fluorophore excitable in the range of 600 to 1000 nm.

34. A method for analyzing a sample of cells or cell constituients having a plurality of detectable characteristics in a container, comprising:
scanning the container with a detector to generate a plurality of channels of data, at least one particular channel containing information relevant to more than one of the plurality of detectable characteristics;
sampling the plurality of channels of data to produce a corresponding plurality of scanned images of the sample; and analyzing the plurality of scanned images to characterize the certain cells or cell constituients in response to the plurality of channels of data, including processing the scanned image corresponding to the at least one particular channel to distinguish the more than one of the detectable characteristics.

35. The method of claim 34, wherein the step of analyzing the plurality of scanned images includes the step of processing at least one of the plurality of scanned images to identify segments of the plurality of scanned images containing certain cells or cell constituients.

36. The method of claim 34, wherein the step of processing the scanned image corresponding to the at least one particular channel includes using another one of the plurality of scanned images in combination to distinguish the more than one of the detectable characteristics.

37. The method of claim 36, wherein the step of using another one of the plurality of scanned images in combination includes performing a linear regression analysis based upon corresponding portions in the at least one particular scanned image and the another one of the plurality of scanned images, to determine a slope value indicating relative contributions of the more than one of the detectable characteristics.

38. The method of claim 35, including filtering the identified segments based upon expected characteristics of the certain cells or cell constituients to generate respective intensity values for the identified segments.

39. The method of claim 34, wherein the container has detectable edges, and at least one of the channels of data includes information relevant to the edges of the container, and including processing at least one of the scanned images to detect the edges of the container; and wherein the step of analyzing the plurality of scanned images to characterize the certain cell or cell constituent includes ignoring pixels in the scanned image outside the detected edges of the container.

40 The method of claim 34, wherein the cells are blood cells.

41. The method of claim 34, wherein the cells are labeled with a fluorophore excitable in the 600 to 1000 nm range.

42. A method for analyzing a sample of cells or cell constituents having two detectable characteristics, comprising:
scanning the sample with a detector to generate two channels of data, each channel containing information relevant to the two of detectable characteristics;
sampling the two channels of data to produce a corresponding plurality of scanned images of the sample; and analyzing the two scanned images to characterize the certain cells or cell constituents in response to the two channels of data, including performing a linear regression analysis over corresponding portions in the two scanned images.

43. The method of claim 42, wherein the step of analyzing the two scanned images includes the step of processing at least one of the two scanned images to identify segments of the two scanned images containing certain cells or cell constituents, and wherein the linear regression analysis is performed over the identified segments.

44. The method of claim 43, including filtering the identified segments based upon expected characteristics of the certain cells or cell constituents to-generate respective intensity values for the identified segments.

45. A method for analyzing in a container with detectable edges a sample of cells or cell constituents having detectable characteristics, comprising:
scanning the container with a detector to generate a channel of data containing information relevant to the detectable characteristics and the edges of the container;
sampling the channel of data to produce a scanned image of the container; and processing the scanned image to detect the edges of the container; and analyzing the scanned image to characterize the certain cells or cell constituents in response to the channels of data, including ignoring portions in the scanned image outside the detected edges of the container.

46. The method of claim 45, wherein the step of analyzing the scanned image includes the step of processing the scanned image to identify segments of the scanned image containing certain cells or cell constituents.

47. The method of claim 46, including filtering portions within the identified segments based upon expected characteristics of the certain cells or cell constituents to generate respective intensity values for certain cells or cell constituents within identified segments.

48. The method of claim 46, including defining a neighborhood of pixels for each identified segment, the neighborhood being larger than an expected size of a certain cell; and processing the pixels within the neighborhood to compensate for background signal and generate an intensity value for the certain cell within the neighborhood.

49. The method of claim 48, wherein the step of processing the pixels within the neighborhood includes filtering pixels within the neighborhood based upon an expected shape of the certain cells to generate an intensity value for a certain cell within the neighborhood.

50. A method for analyzing a sample of cells or cell constituents with detectable characteristics in a container having edges, comprising:
scanning the container with a detector to generate a channel of data containing information relevant to the detectable characteristics and the edges of the container;
sampling the channel of data to produce a scanned image of the container;
processing the scanned image to identify segments of the scanned image containing certain cells or cell constituents; and filtering portions within the identified segments based upon expected characteristics of the certain cells or cell constituents to generate respective intensity values for certain cells or cell constituents within identified segments.

51. The method of claim 50, wherein the step of filtering includes defining a neighborhood of pixels for each identified segment, the neighborhood being larger than an expected size of a certain cell; and processing the pixels within the neighborhood to compensate for background signal and generate an intensity value for the certain cell or cell constituent within the neighborhood.

52. The method of claim 51, wherein the step of processing the pixels within the neighborhood includes filtering portions within the neighborhood based upon an expected shape of the certain cells or cell constituents to generate an intensity value for a certain cell or cell constituent within the neighborhood.

53. A method for analyzing a sample of cells or cell constituents, the cells or cell constituents having detectable characteristics, comprising:
scanning the sample with a detector to generate a channel of data containing information relevant to the detectable characteristics;
sampling the channel of data to produce a scanned image of the sample;
processing the scanned image to identify segments of the scanned image containing certain cells or cell constituents;
defining a neighborhood of pixels for each identified segment, the neighborhood being larger than an expected size of a certain cell; and saving the neighborhood of pixels defined for each identified segment, for further processing.

54. The method of claim 53, wherein the further processing comprises:
processing the pixels within the neighborhood to compensate for background signal and generate an intensity value for the certain cell or cell constituent within the neighborhood.

55. The method of claim 54, wherein the step of processing the pixels within the neighborhood includes filtering pixels within the neighborhood based upon an expected shape of the certain cells or cell constituent and background intensity to generate an intensity value for a certain cell or cell constituent within the neighborhood.

56. A method for analyzing a sample of cells or cell constituents, the cells or cell constituents having detectable characteristics, comprising:
scanning the sample with a detector to generate a channel of data containing information relevant to the detectable characteristics;
sampling the channel of data to produce a scanned image of the sample;
processing the scanned image to identify segments of the scanned image containing certain cells or cell constituents;
defining a neighborhood of pixels for each identified segment, the neighborhood being larger than an expected size of a certain cell or cell constituent; and saving a set of cell or cell constituent parameters calculated from the neighborhood of pixels defined for each identified segment, for further processing.

57. The method of claim 56, wherein the further processing comprises:
processing the pixels within the neighborhood to compensate for background signal and generate an intensity value for the certain cell or cell constituent within the neighborhood.

58. The method of claim 57, wherein the step of processing the pixels within the neighborhood includes filtering pixels within the neighborhood based upon an expected shape of the certain cells or cell constituent and background intensity to generate an intensity value for a certain cell or cell constituent.