Particle sorter

Abstract

A particle sorter for sorting particles being conveyed in a fluid stream, the sorter comprising: a particle detection system, for detecting particles in the stream; a particle deflection system, for selectively deflecting particles from the stream to sort them; an optical imaging system arranged to capture images of particles within the stream so as to identify particles at different locations within the stream; an illuminating light source for selectively illuminating the stream when capturing images of particles; and an electronic control system, wherein the sorter is configured such that, in use, when the particle detection system signals a detection event, the control system: controls the imaging system and the illuminating light source so as to capture one or more images of the particle in the stream with a respective one or more time delays from the detection event; and processes the one or more images and the respective one or more time delays to determine one or more parameters of the sorting process.

Description

[0001] PARTICLE SORTER

[0002] The present invention relates to a particle sorter.

[0003] BACKGROUND

[0004] Instruments for particle sorting have widespread uses in biological research. A major application of particle sorting technology in this field is to sort biological cells. Instrumentation for sorting cells based on measurements of fluorescent labels within the cells is typically known as fluorescence activated cell sorting (FACS). Other applications of particle sorting include the sorting of solid beads or liquid droplets of one liquid phase in a carrier fluid. For example, aqueous droplets in a non-aqueous carrier fluid can be used to contain cells.

[0005] In a conventional particle sorter, the detection of individual particles in a flowing stream is followed by the generation of a jet through a nozzle, this jet is projected through the air and broken into droplets that are then electrostatically steered to one or more output vessels. One major disadvantage of conventional particle sorters when used for sorting of biological cells is that the cell suspension is exposed to the air and to non-sterile surfaces. Thus the particle sorter may expose the user to biohazards via escape of aerosols. The particle sorter may also allow contamination of the cell suspension by microbes on non-sterile surfaces, or crosscontamination of the cells via carry-over of previous samples in the same wetted parts. To allow sterile cell sorting, and for many other potential benefits, there are various kinds of microfluidic particle sorters. Like a conventional sorter, the microfluidic sorters measure individual cells in a stream. However, they employ a variety of mechanisms to deflect a target particle.

[0006] All of the above particle sorting devices usually comprise a means to measure individual particles in a flow, an electronic control system able to discriminate (“gate”) particles into target and non-target populations in real time based on these measurements, and a deflection mechanism that is sufficiently precise to deflect an individual particle, if the mechanism is actuated when the particle is at a certain point in the flow (referred to as “actuation location” or “deflection location”). In any of the above cell sorting devices, an “actuation delay time”, defined as the interval between the detection event and the actuation of the deflection mechanism, must be set in the electronic control system. This actuation delay time must accurately equal the passage time that a target particle takes to travel between the detection location and the actuation location. In order to determine the passage time and therefore set the actuation delay time, which is required for all of the above types of particle sorter, the user must pass fluorescent tracer beads through the system before a cellular sample is flowed. A region downstream of the deflection is then observed, typically with a bright illumination and an electronic video camera with a sufficiently long exposure time to image the average path of fluorescent particles. The actuation delay time is then scanned until the fluorescent particles are deflected into the desired path.

[0007] The disadvantages of this method of setting the actuation delay time are firstly that the passage time may be affected by many physical variables, such as the velocity of the particles within the stream, the temperature and viscosity of the medium, and the detection and actuation locations. Even if the actuation delay time is set immediately prior to running the biological sample, many of these effects may rapidly cause a variation of passage time such that the performance of the sorter quickly deteriorates. If the actuation delay time does not equal the passage time, then the target particle will not be deflected, causing a loss of purity and yield in the sorting process.

[0008] As can be seen from the above, there is a need to improve the accuracy of the sorting process whilst preferably also simplifying the manufacture and set-up of the particle sorter. It is also desirable to avoid the need to employ tracer beads.

[0009] SUMMARY OF THE INVENTION

[0010] The present invention provides a particle sorter for sorting particles being conveyed in a fluid stream, the sorter comprising: a particle detection system, for detecting particles in the stream; a particle deflection system, for selectively deflecting particles from the stream to sort them; an optical imaging system arranged to capture images of particles within the stream so as to identify particles at different locations within the stream; an illuminating light source for selectively illuminating the stream when capturing images of particles; and an electronic control system, wherein the sorter is configured such that, in use, when the particle detection system signals a detection event, the control system: controls the imaging system and the illuminating light source so as to capture one or more images of the particle in the stream with a respective one or more time delays from the detection event; and processes the one or more images and the respective one or more time delays to determine one or more parameters of the sorting process. In the particle sorter of the invention, a sample of individual detection events trigger the camera and illumination with a precise time lag (also referred to as a time delay) starting from the event detection. The position of the particles within the camera frame can be detected automatically using a machine vision algorithm. Thus in each frame, the precise particle location is measured at a precise time lag.

[0011] The particles to be sorted may be, for example, cells, beads, or droplets containing further particles. In this document, the terms “cell sorter” and “particle sorter” are used interchangeably, by which we mean a device that can be used to sort any or all of the above particles suspended in a carrier fluid. The particle sorter may be a microfluidic particle sorter.

[0012] The deflection of the particles within the sorter may be generated in a number of ways, for example by employing acoustic waves, electrophoretic forces, optical forces, magnetic forces, microscopic valves, or hydrodynamic forces via fast transducers such as piezoelectric or thermal vapour bubble actuators.

[0013] The particle sorter of the invention may set the actuation delay time (defined as the interval between a detection event and the actuation of the deflection mechanism) as follows. The value of the time lag can be adjusted until the particle is at the actuation location. At this point the time lag equals the passage time. The actuation delay time is then set equal to this time. For this first function, the imaging system’s field of view preferably encompasses at least the actuation location. The passage time that a target particle takes to travel between the detection location and the actuation location is an example of a measured parameter of the sorting process. The actuation delay time is an example of a control parameter which is adjusted for optimal operation of the particle sorter.

[0014] The particle sorter of the invention may trigger a single flash to illuminate the fluid stream after a time delay and may trigger a camera frame to be captured after a camera time delay, both synchronised with a corresponding particle detection event. The camera time delay may be chosen so that the camera shutter is open at the moment that the illumination takes place. The time delay may be varied to capture various information about the particle being sorted. The time delay may be set equal to the actuation time delay, so that the particle location can be measured in an image at the moment of actuation. Typical flash duration is currently 0.8 microseconds, in general it should be of the order of a microsecond or shorter

[0015] A second function of the particle sorter may be to adjust the actuation amplitude. Here the value of the time lag is increased so that the particle is imaged at a point that its output path is visible in the image. The actuation amplitude can then be adjusted until the particle deflection corresponds to a target flow path. For this second function, the imaging system’s field of view preferably encompasses at least one location downstream of the actuation location, after the deflection of the particle has occurred.

[0016] A third function of the particle sorter of the invention may be to count the proportion of sampled detection events for which the particle is deflected within a certain range of amplitude such that it flows into the correct output (along a target flow path). We use the term “fidelity” for this proportion. It is useful to measure fidelity in the cell sorter automatically and report this to the user, since the fidelity gives a basic check that the sort process will give a high-purity, high- yield output.

[0017] A fourth function of the particle sorter of the invention may be to count the proportion of sampled detection events for which the particle is present at the precise actuation location. We use the term “actuation presence” for this proportion. It is useful to measuring actuation presence in the cell sorter automatically and report this to the user, since it gives a basic check that particle detection events correspond to a precise time and location and that the velocity of the particle between detection and actuation locations is well controlled.

[0018] A fifth function of the particle sorter of the invention may be to count the frequency of particles present in the imaged deflected path that are not synchronized with a deflection event. This frequency divided by the time sampled in the deflected path equals the “output error rate”. It is useful to measure output error rate automatically and report this to the user, since it gives a basic check that there are no failures of particle focusing that would adversely affect yield and purity.

[0019] A sixth function of the particle sorter of the invention may be to measure the standard deviation of particle positions perpendicular to the flow direction at the actuation location. We define this measure as the “particle focusing quality”. It is useful to measure particle focusing quality automatically and report this to the user, since it gives a further basic check that there are no failures of particle focusing that would adversely affect yield and purity.

[0020] A seventh function of the particle sorter of the invention may be to measure the standard deviation of particle positions parallel to the flow direction at the actuation location. We define this measure as the “particle position jitter”. It is useful to measure particle position jitter automatically and report this to the user, since it gives a basic check that there are no failures of particle detection timing that would adversely affect yield and purity. An eighth function of the particle sorter of the invention may be to measure the arrival and build-up of debris within the particle sorter. We define several measures depending on the location where the debris arrives. Each measure is the area of the camera frame occupied by debris divided by the relevant area of the location. For example the debris at the actuation location or at a downstream location where the flow channel splits are defined as “actuation location gunk proportion” and “junction gunk proportion”. It is useful to measure automatically and report gunk proportions to a user since it gives a basic check that the sample preparation is sufficient to avoid blocking of the particle sorter. It is also useful to measure gunk automatically since the sorter may be automated to unblock the channel by changing the flow path if gunk is detected.

[0021] Examples of sorting process parameters include: particle position relative to actuation location, measured particle deflection magnitude, presence of particle in target flow path, time of particle passage from detection location to an actuation location, fidelity, actuation presence.

[0022] Examples of control parameters include: actuation time delay, actuation amplitude, illumination (e.g. strobe) time delay, camera time delay. Furthermore, control parameters can include one or actuation control parameters which control how the deflection system operates - for example defining a particular voltage waveform in order to achieve a target deflection of the particle, defining a simple magnitude and / or duration of a voltage signal, or in the case of acoustic wave actuation defining an amplitude and / or duration of the acoustic wave.

[0023] FIGURES

[0024] Figure 1 shows the schematic layout of prior art particle sorters in which figure 1a shows a generic particle sorter describing both conventional particle sorters and microfluidic particle sorters, figure 1 b shows a conventional particle sorter, and 1 c shows a vortex-actuated particle sorter.

[0025] Figure 2 shows a schematic of the particle sorter optical instrumentation system employed in the present invention.

[0026] Figure 3 shows particle sorter optical instrumentation layout according to a first example of the invention. Figure 4 shows a schematic of an example particle sorter control system according to the invention.

[0027] Figure 5 shows a particle sorter control system according to a first example of the invention.

[0028] Figure 6 shows a schematic of a particle sorter machine vision algorithm capable of identifying particles in the camera frame that can be employed in the invention.

[0029] Figure 7 shows a particle sorter machine vision algorithm capable of identifying particles in the camera frame according to the first example of the invention.

[0030] Figure 8 shows example results from using a particle sorter according to the invention to set actuation delay time and actuation amplitude.

[0031] Figure 9 shows the cell sorter machine vision detection masks within the camera frame according to the first embodiment.

[0032] Figure 10 shows example results reporting the functional measurements to a user via a graphical user interface.

[0033] DETAILED DESCRIPTION

[0034] Referring to the figures, figure 1 shows prior art sorting configurations. The schematic of a generic particle sorter is shown in Fig 1a. In this device the stream of particles is focused in a flow path 101 , that passes through a detection zone 102 which comprises one or more detection locations 103, 104, 105 where the particle is measured. The passage of the particle through each detection location gives a time-resolved signal to the particle detection system. If there are more than one detection locations, the signal from the upstream detection locations 103, 104 is matched with the signal from the downstream detection location 105 to give an event. An event is the passage of a particle which occurs at a certain time at a certain detection location, by which point a set of measurements have been made on the particle, which may comprise fluorescence measurements, light scatter measurements and time measurements at one or more detection locations. The event is processed by an electronic control system, which decides whether or not the particle belongs to one or more target gates based on the set of particle measurements, and therefore whether to sort the particle into an output stream that corresponds to each target gate. For example, the generic particle sorter of Fig 1a has two output streams 108 that correspond to two target gates (a “two-way sorter”), as well as a waste stream 109 for non-target particles. If a decision is made to deflect a particular particle, the deflection is made by a deflection mechanism 106, which acts on a particle at the actuation location 107, which is a particular location through which the particle passes, downstream of the detection location. The electronic control system is programmed so that the deflection mechanism is actuated at a precise actuation delay time after the event, with a precise actuation amplitude so that the particle enters the required output stream. The actuation delay time and actuation amplitude are parameters required by the electronic control system, and both must be determined accurately. The actuation delay time must be set equal to the passage time of target particles between the detection location and actuation location, otherwise the target particle will not be sorted. The actuation amplitude must be set so that particles of a particular target gate receive the deflection to send them into the corresponding output stream. As described above, a lack of accuracy in these control parameters can very quickly result in inefficient sorting and poor operation.

[0035] In conventional particle sorters, the actuation location is the point at which a droplet breaks off from the stream. The solution to the timing problem is firstly to have a stable particle velocity (requiring stable pressure and temperature of the instrument and the flowing medium), and secondly to have a stable location of droplet break-off (requiring stable stimulation of the stream). The droplet break-off is typically monitored using stroboscopic video imaging. The stream is imaged onto a video camera at the point that it breaks into droplets. The illumination is pulsed regularly with a frequency equal to the droplet stimulation frequency. The phase of the illumination pulses is then scanned to determine the precise location of droplet creation, and to check that this location does not move. In the typical set-up, the camera exposure is much longer than the stimulation period. Thus many illumination pulses are captured in a single camera frame, in order to give an image of the droplet creation averaged over many stimulation cycles. In this way, the stroboscopic imaging system determines the actuation location as we have defined it above. However, this stroboscopic system cannot determine the actuation delay timing, for two reasons. Firstly, the velocity of the particles within the stream is not known and not possible to measure by this method, since the particles themselves cannot be seen within the droplets. Secondly, the strobe illumination is synchronized with the droplet stimulation oscillation, not with the detection event. The detection events are indeed not correlated to the phase of the droplet stimulation oscillation.

[0036] Another prior art particle sorter is shown in Fig 1b and follows the scheme of the particle sorter of figure 1a. The flow enters a cuvette 111 at elevated pressure. The particles are focused by hydrodynamic focusing (upstream of this diagram; not shown) into a flow path 112, which takes them through a series of laser foci 113, which are the detection locations described above. A set of fluorescence, light scatter and time measurements are made on each particle. The particle suspension exits through a nozzle 114 in a jet 115 which breaks into droplets. The fluid is acoustically stimulated so that each droplet breaks off at a precise phase of the stimulation cycle. The location 116 at which a droplet breaks off is the actuation location above. At a precise actuation time delay after the detection event, the control system applies a precise voltage to an electrode 117 in contact with the particle suspension upstream of the nozzle. This voltage is the actuation amplitude in this example. The voltage causes an electric polarization of the jet so that, at the moment of break-off, the droplet is charged electrostatically in proportion to the voltage. The droplet 120 then flies towards a set of output receptacles 121 while a static electric field between two charged plates 118 cause the deflection of the droplet in proportion to the trapped charge on the droplet. One stream (“waste”) is designated to receive the non-target particles 119. The actuation amplitude must be further adjusted to correct for electrostatic interactions between a current droplet and the preceding droplet. It is important that the speed of the jet and the location of droplet break-off are stable, since any drift of either of these would render the actuation delay time unequal to the passage time, and reduce the accuracy of the sorting.

[0037] Another microfluidic particle sorter is shown in Fig 1c and also follows the scheme of the particle sorters of figures 1a and 1 b. The flow takes place in a microfluidic channel on a chip 131. Particles enter in a focused stream 132 which passes through one or more laser foci (133, 134) at precise points within the microfluidic channel 135: these laser foci are the detection locations described above. Measurements of a particle from all laser foci are combined in the electronic control system so that the event of each particle passing the downstream laser focus 133 is accompanied by the set of particle measurements (fluorescence, scatter and time measurements from all laser foci). At an actuation delay time after the event, the electronic control system then applies a voltage pulse across the microresistor 136. The voltage pulse is defined by an amplitude (the actuation amplitude) and a time interval. This voltage pulse heats up the resistor so that a film of superheated liquid in contact with the resistor forms a thermal vapour bubble. The thermal vapour bubble rapidly expands within a few microseconds, causing a transient flow across the vortex-generating tip 137 in the channel wall, which causes a transient vortex to form in the flow downstream of this tip. If particle is at the actuation location 138 close to the tip 137 when the actuation occurs, it will flow downstream at the same time as the vortex, causing its deflection from an undeflected flow path 139 to a deflected flow path 140. These two paths enter two corresponding outputs 142 and 143 downstream of a junction 141 defined by the wall of the microfluidic channel. We refer to the two output channels as “Sort 0” (undeflected) and “Sort 1” (deflected). While Fig 1c shows a device with a single output 143 for the deflected particles (a “single-way sorter”), this type of sorter can accommodate several output channels branching off the input channel. The voltage of the actuation pulse determines the magnitude of the transient vortex, which varies the amplitude of deflection perpendicular to the streamline of the undeflected particles. Therefore this deflection mechanism can send an particle towards a choice of output channels (in embodiments comprising more than one output channel). Whilst this provides accurate and reliable sorting it still has a requirement for high control parameter accuracy to maintain this.

[0038] Figure 2a shows a schematic of particle sorter optical instrumentation employed in the present invention. A flow channel 250 carries a fluid stream in which particle s are conveyed. The flow channel 250 is illuminated by one or more lasers, and at least one of forward scattered light, occultation, side scattered light and fluorescence are measured. (Occultation is the loss of light from the undeflected beam due to scattering and absorption. Occultation is commonly also referred to as “forward scatter” in the art, since it is approximately a constant signal of the direct beam minus the forward scattered light and therefore has approximately the same information as true forward scattered light.) The measurement of at least one of forward scattered light, occultation, side scattered light and fluorescence is performed by a particle detection system in order to detect a particle in the fluid stream of the flow channel.

[0039] An illumination light source is provided and a camera set up to image at least part of the flow channel 250. The instrumentation of illumination and measurement may be combined in various ways using lenses, filters and dichroic mirrors. For example, the illumination source and camera may share an objective lens (epi-illumination), which may be further combined with the laser illumination, the side-scattered light, fluorescence detection or forward scattered light.

[0040] The illumination and camera may be of the strobe type, although it will be appreciated that this is not essential. Advantageously, and as explained further below, the illumination source and camera can be controlled to capture a camera frame where the camera shutter is open at the moment that the illumination (e.g. strobe flash) takes place. The camera shutter preferably only remains open for a duration to capture a single illumination, in order that a target particle is imaged. In some cases, multiple particles may be imaged, including the target particle.

[0041] Fig 2b shows a more detailed schematic of particle sorter optical instrumentation according to the invention. One or more lasers 201 are each focused and shaped by one or more lenses 202, and combined by a series of dichroic mirrors 203, after which they are combined by projecting lens 204, which focuses them into flow channel 205 that may be contained in a cuvette or chip. If the flow channel contains a reflective surface, a forward collection lens 206 is positioned to collect the specularly reflected light. If the flow channel is transparent, the forward collection lens is positioned to collect the transmitted light (not shown). Either way, the collimated light is split by a series of dichroic mirrors 207. Each beam is then focused by a lens 208 onto a detector 209. Optionally a beam stop 240 is placed to stop the unscattered light of each beam. If the beam stop is present, the detector measures forward scattered light. If the beam stop is not present, the detector measures occultation. An objective lens 210 projects illumination light onto the flow channel and collects side scattered light, fluorescence emission and illumination reflection from the flow channel. Dichroic 211 splits the illumination light from the laser excitation light. The illumination wavelength is chosen so that it is well separated from the fluorescence emissions that the instrument is set up to detect. The side scattered light and fluorescence light passes through the dichroic and is split by a further series of dichroic mirrors 212 into side scattered light and one or more fluorescence emission bands. Each emission band is focused 213 onto a light detector 214, which the side scattered light is similarly focused 215 onto a detector 216. The illumination light comes from light source 218 (which may be a strobe light source) and is collimated by lens 219. Partially silvered mirror 217 sends the illumination light to the chip via dichroic 211 and objective lens 210, while allowing the reflected light to return by the same path. Some of the reflected light passes through 217 and is focused by lens 220 onto the camera 221.

[0042] Fig 3 shows a particle sorter optical instrumentation layout according to the invention. All parts are attached to a base plate 300. A microfluidic chip 301 is mounted such that the incident laser beams 302 are focused on the flow channel at the detection locations described above (chip mount not shown). The fluorescence emission, side scattered light and imaging 303 from the chip is collected separately from the specular reflection 304 as described below. The laser projecting block 305 takes two collimated, shaped laser beams (488 nm “blue” and 640 nm “red beams) as an input (laser modules and beam shaping stages not shown): the beams are steered and combined by movable mirrors 306 and fixed dichroic mirrors 307, before being projected onto the chip by lens 308. The forward scatter collection block is labelled 309. The specularly-reflected laser beam exits the chip and enters the forward-scatter collection lens 310, thence is split into red and blue components by dichroic mirror 312, thence one beam is further reflected by mirror 311 , so that the specularly reflected beams are separated and collected onto photodiode detectors 313. The block 314 houses fluorescence collection, side scatter, illumination and imaging. The objective lens 315 projects illumination onto the chip and collects and collimates side-scattered light, fluorescence emissions and illumination reflection from the chip. The collimated light is split by a series of dichroic mirrors 316 and a knife-edge pick-off mirror 322 into fluorescence bands that are measured by a set of silicon photomultiplier detectors 324. The side scattered light is measured by photodiode 325. The illumination component in this example employs an LED 318 at a wavelength (850 nm) chosen to be far from the laser wavelengths and the fluorescence emissions. The light is projected onto the chip, and the reflected light returns to the same path through the dichroic mirrors since its wavelength is not altered. This light is split by partially silvered mirror 317, after which it is focused by tube lens 320 onto the video camera 321.

[0043] Figure 4 shows a schematic of a particle sorter control system according to the invention. One or more measurement channels (“n channels”, which may comprise forward scattered light, occultation, side scattered light, one or more fluorescence bands) are input to a signal processing system. The signal processing system detects events within the measurement channels: each event is characterised by a detection time and a set of measurements according to the n measurement channels. Each event may include information from a particle passing through more than one laser at different times: this information must then be spliced to identify the same particle in the separate signals. The events are fed into a sort logic system which applies a set of threshold criteria (“gates”) to discriminate target events from non-target events, and decide which target events to sort into which output. For each event that is decided to be sorted (“sort events”), a trigger signal is sent to the actuator (e.g. the particle deflection system) after an actuation time delay (“actuation trigger”). For at least some of these sort events, the decision is also made to capture a camera frame (“strobe events”). In each strobe event, a strobe trigger is sent after a strobe time delay and a camera trigger after a camera time delay, both synchronised with the corresponding sort event. The camera time delay is chosen so that the camera shutter is open at the moment that the illumination takes place. The strobe time delay may be varied to capture various information about the sort event. In the first function, the strobe time delay is set equal to the actuation time delay, so that the particle location can be measured in an image at the moment of actuation. The cases of further functions are detailed below.

[0044] Figure 5 shows an example particle sorter control system according to the invention. Two measurement channels are provided, each from a separate laser (Laser 1 and Laser 2), both fed into analog to digital converters (ADCs), thence entering a field programmable gate array (FPGA). The Laser 2 channel is delayed in a buffering block by a Laser Delay parameter, and both signals are fed into signal processing blocks, each of which include a finite impulse response (FIR) filter, baseline correction (BLC correction) and peak detection. Each signal processing block sends a stream of event records to a peak matching block, which splices the records from the two laser channels. Each spliced event record represents a single particle, where a time difference (“DeltaT”) is calculated between the signals in the two lasers. The spliced event record enters a further processing block, which performs cytometric compensation, gating and sort logic, so that a decision is made whether to sort or not sort each event. The event records are sent to a further block which generates an actuation pulse trigger at an actuation delay time, and selects a subset of events for imaging. Several parameters of the sorting process are measured from the imaging as described below, and used to update the control parameters such as the actuation delay time and illumination delay time in the FPGA, for various functions as described below. This block also sends the stream of events to a microcontroller unit (MCU) for determination of the Laser Delay parameter. For each event to be imaged, the block generates a camera trigger and a strobe trigger. The strobe trigger activates an LED flash of around 1 microsecond. The camera trigger is timed to be a few microseconds before the strobe trigger so that the camera shutter is open when the LED flash occurs. In the MCU, the event data is analysed automatically to update Laser Delay by the following method. A median DeltaT is calculated over a certain time period. This median is a measure of the error in the Laser Delay, which is used to calculate a new Laser Delay. The new Laser Delay is fed back to the FPGA buffering block. The camera image is fed to an embedded PC for automatic image analysis. Image analysis in the embedded PC determines the position of the particle at the actuation delay time. This position is fed to the Control Systems block, which determines several adjustments. In the first function, the Control System calculates what adjustment should be made to the actuation delay time to bring the particle to the actuation position.

[0045] Figure 6 shows a schematic of a particle sorter image analysis and control system according to the invention. Particles are first identified in each image by a particle detection algorithm. Then the particle positions are correlated to the event times. It will be appreciated that various methods of correlation are possible. In a direct method with indexed images (where each image is associated with an indexed event), events are taken that are well spaced from their neighbours, so that there is only one particle in the relevant region of each image. Thus there is an unambiguous association of particle position with illumination time delay in a single image. The particle position is then fed to a controller, which decides an increment of the actuation time delay based on an actuation position set point. Following this decision, the illumination delay is set to the new actuation time delay; new sort events are triggered and new images acquired, which are then fed back into the particle detection algorithm, thus completing the control loop.

[0046] Figure 7 shows particle sorter image analysis capable of identifying particles in the camera image according to the invention, and the associated control loop. The particle detection algorithm comprises several steps. (1) Background subtraction: the absolute difference is taken between the new image and a background images without particles. This image is then thresholded (2) and subjected to noise reduction (3) by morphological opening (erosion followed by dilation). This processed image is fed to a clustering algorithm (4), for which we use the known algorithm DBScan, producing a set of particles in the image. These particles are then filtered (5) by size, weight, density and circularity to eliminate image artifacts. The filtered particle positions are correlated by an accumulator algorithm which takes the particle position of several frames in order to produce an unambiguous particle position corresponding to the illumination time delay. The accumulator algorithm overcomes cases where there are many particles per frame and ambiguity about which event corresponds to which particle image. Various accumulator algorithms are possible: the one employed in this example consists of putting each particle position as a weight at one location of a one-dimensional position array, this location corresponding to the distance along the contour line that a particle takes through the image in the sorter (“streakline”). Particles from several (“n”) frames are accumulated on this array to build up a certain weight. This array is then convolved with a smoothing function, for which we take the sum of a positive Gaussian (with width equal to the expected uncertainty of particle positions caused by a single illumination time delay), and a negative Gaussian (with a greater width corresponding to the likely positions of the nearest neighbouring particles). This function is therefore more likely to find the position corresponding to target events, since they have the same illumination time delay, and less likely to find neighbouring particles. A proportional controller then calculates the difference between the particle position and the actuation position set point, and decides an actuation position increment as a proportion of this difference. More generally PID (proportion-integration- differential) controllers may be used, but we have found that a proportional controller is sufficient. The actuation position increment is fed to a velocity correction function, which calculates an actuation time delay increment based on the measured speed of the particles in the stream by their passage between the two lasers. The illumination delay is then set to the new actuation time delay, new sort events trigger the acquisition of new images, which are then fed back into the particle detection algorithm, thus completing the loop.

[0047] The second function of the automated particle sorter, to adjust the actuation amplitude, follows the description of the particle detection and control loop above, with a few modifications. In this second function, particle positions are determined in a downstream region after the deflection has occurred. The strobe delay time is set to a “verification delay time” instead of the actuation delay time. The correlation algorithm measures the particle deflection from the undeflected particle path. The controller takes a set point deflection and calculates an increment of the actuation amplitude to reach this set point. Figure 8 shows example results of using the particle sorter of the invention to set the actuation delay time and actuation amplitude. 801 shows an image triggered at the actuation time delay, with a detail shown 802. Overlaid on the image is a particle detection region around the actuation location (“actuation mask”) 804, and a graphical output of the accumulator algorithm 803, showing both the weight histogram (bars) and the convolved function (line graph) along the streakline. The particle is detected within the image 805, and the actuation location set point shown 806. A further image frame is shown 807 triggered at the verification time delay. Overlaid on this image is the particle detection region downstream of the particle deflection (“verification mask”) 808, showing the deflected particle 809.

[0048] Using the particle sorter control system and image analysis algorithm as described above, several further functions are enabled. The third function of the particle sorter is to measure the fidelity of the sorter, i.e. the proportion of images, triggered at the verification time delay, for which the particle is deflected within a certain range of amplitude such that it flows into a particular deflected output channel instead of the undeflected output channel. In the first example, these output channels are the Sort 1 and Sort 0 channels described above. The fourth function of the particle sorter is to measure the actuation presence, i.e. the proportion of images, triggered at the actuation time delay, for which the particle is present at the precise actuation location. The fifth function of the particle sorter is to measure the output error rate, i.e. the frequency of particles present in the sampled images, triggered at the actuation time delay, in which particles are present in the deflected output channel, which are thus not synchronized with a deflection event. In the first embodiment, the output error rate may be expressed as the expected rate of false positives in the Sort 1 output, which accounts for the expected rate of asynchronous coincidence events. I.e. the output error rate equals the proportion of frames in which a deflected particle was detected in the Sort 1 output asynchronously with respect to the actuation time delay, divided by the time in which a flowing particle is expected to be in the imaged region of the Sort 1 output, minus the sort rate. The sixth and seventh functions of the particle sorter are to measure the particle focusing quality and particle position jitter respectively. These are defined as the standard deviations of the particle positions at the actuation point in “y” (the vertical axis, perpendicular to the streakline) and “x” (the horizontal axis, parallel to the streakline), respectively, of the images triggered at the actuation delay time, with axes defined according to the image shown in Figure 8. The eighth function of the particle sorter is to measure gunk at several regions of interest (“gunk masks”) within the image field-of-view.

[0049] Gunk is measured as the proportion of each gunk mask that has a significant difference from a reference image without gunk. The gunk measurement is used in the particle sorter to trigger a separate mechanism of gunk clearing. In the first embodiment, the mechanism of gunk clearing consists of the transient blocking of the Sort 1 output by a solenoid valve outside of the sorter chip, and optionally the application of a larger amount of electrical energy to the actuation than is required for actuation of deflection.

[0050] Figure 9 shows the particle sorter machine vision detection masks within the camera frame according to the invention. Each mask is a region of interest for the image analysis algorithms described. 901 shows the cell sorter’s microfluidic channel outline. 902 shows the actuation mask. 903 shows the verification mask. Three gunk detection masks are shown at the detection region 904, the actuation region 905, and the junction between Sort 1 and Sort 0 output channels 906.

[0051] Figure 10 shows example results reporting the functional measurements to a user via a graphical user interface according to the invention. The particle position jitter, particle focusing quality, gunk proportion (at the junction region), actuation presence, deflection, fidelity and output error rate are indicated.

[0052] As will be appreciated from the above, the present invention can improve the operation and accuracy of particle sorting by identifying parameters in the process using relatively simple image capture equipment. The employment of a delayed triggering illumination and image capture configuration allows the use of relatively inexpensive and simple to configure equipment.

Claims

CLAIMS1. A particle sorter for sorting particles being conveyed in a fluid stream, the sorter comprising: a particle detection system, for detecting particles in the stream; a particle deflection system, for selectively deflecting particles from the stream to sort them; an optical imaging system arranged to capture images of particles within the stream so as to identify particles at different locations within the stream; an illuminating light source for selectively illuminating the stream when capturing images of particles; and an electronic control system, wherein the sorter is configured such that, in use, when the particle detection system signals a detection event, the control system: controls the imaging system and the illuminating light source so as to capture one or more images of the particle in the stream with a respective one or more time delays from the detection event; and processes the one or more images and the respective one or more time delays to determine one or more parameters of the sorting process.

2. The particle sorter of claim 1 , wherein the control system is further configured to adjust at least one control parameter for the particle sorter based on the one or more determined sorting process parameters.

3. The particle sorter of claim 2, wherein the adjusted control parameter is one or more of: the time delay for capturing the one or more images, a time delay for activating the illuminating light source in order to capture the one or more images, an actuation control parameter which controls actuation of the particle deflection system, and a timing and / or magnitude of the deflection of the particle by the particle deflection system.

4. The particle sorter of any preceding claim, wherein the one or more sorting process parameters are determined by identifying the position of one or more particles in the captured one or more images.

5. The particle sorter of any preceding claim, wherein the control system processes the captured one or more images to count the proportion of sampled detection events forwhich the particle is deflected within a certain range of amplitude such that it flows into the correct output.

6. The particle sorter of any preceding claim, wherein the control system processes the one or more captured images to count the proportion of sampled particle detection events for which the particle is present at an actuation location where the particle is deflected by the particle deflection system.

7. The particle sorter of any preceding claim, wherein the control system processes the one or more captured images to determine the frequency of particles present in a deflection path that are not synchronized with a deflection event.

8. The particle sorter of any preceding claim, wherein the control system processes the one or more captured images to determine the standard deviation of particle positions perpendicular to the flow direction at an actuation location where the particle is deflected by the particle deflection system.

9. The particle sorter of any preceding claim, wherein the control system processes the one or more captured images to determine the standard deviation of particle positions parallel to the flow direction at the actuation location.

10. The particle sorter of any preceding claim, wherein the control system processes the one or more captured images to determine the arrival and build-up of debris within the particle sorter.

11. The particle sorter of any preceding claim, wherein the optical imaging system and illuminating light source are arranged to capture the one or images of the particle in the stream when the particle is at an actuation location where the particle is deflected by the particle deflection system.

12. The particle sorter of any preceding claim, wherein the control system is configured to control the optical imaging system and illuminating light source to apply a single flash of light to the stream during the capture of an image.

13. The particle sorter of any preceding claim, wherein the particle sorter comprises a microfluidic particle sorter.

14. The particle sorter of any preceding claim, wherein the particle sorter comprises a single-junction sorter, wherein particle sorter comprises an input channel, one or more output sort channels and an output waste channel, and wherein each of the one or more output sort channels and the output waste channel are connected to the input channel for receiving the fluid therefrom, the particle deflection system configured to selectively deflect particles into the one or more output sort channels or into the output waste channel.

15. The particle sorter of any preceding claim, wherein the sorter is further configured such that, in use, when the particle detection system signals a further detection event, the control system: controls the imaging system and the illuminating light source so as to capture one or more further images of the further particle in the stream with a respective one or more further time delays from the further detection event; and processes the one or more further images and the respective one or more further time delays to adjust the one or more parameters of the sorting process.

16. A method of sorting particles comprising the steps of: supplying a fluid stream to a particle sorter according to any preceding claim; detecting particles in the stream and deflecting the detected particles.

17. The method of claim 16, wherein the particles in the stream comprise particles to be sorted and the step of deflecting the detected particles comprises sorting the particles.

18. The method of claim 16 or claim 17, wherein the particles comprise at least one of: cells, beads, or droplets containing further particles.

Images