CN109859188A - A kind of fluorescence cross-talk correction method and its application based on mean shift algorithm - Google Patents
A kind of fluorescence cross-talk correction method and its application based on mean shift algorithm Download PDFInfo
- Publication number
- CN109859188A CN109859188A CN201910099554.4A CN201910099554A CN109859188A CN 109859188 A CN109859188 A CN 109859188A CN 201910099554 A CN201910099554 A CN 201910099554A CN 109859188 A CN109859188 A CN 109859188A
- Authority
- CN
- China
- Prior art keywords
- data
- fluorescence
- channel
- value
- point
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Landscapes
- Investigating, Analyzing Materials By Fluorescence Or Luminescence (AREA)
Abstract
The present invention relates to machine learning techniques field more particularly to a kind of fluorescence cross-talk correction methods and its application based on mean shift algorithm for this.The fluorescence cross-talk correction method based on mean shift algorithm is the following steps are included: S1: the pretreatment of multichannel fluorescence intensity data；S2: different channel cluster centres are calculated based on mean shift clustering；S3: calculating slope, fills correction matrix, obtains revised fluorescence data.The human error that bearing calibration disclosed by the invention introduces when can be to avoid sampling fluorescence intensity level, noise is reduced to the element value of fluorescence intensity influenced and can quickly and accurately calculate correction matrix, to obtain revised fluorescence data, obtained fluorescence data is truer and more acurrate.
Description
Technical field
The present invention relates to machine learning techniques field more particularly to a kind of fluorescence crosstalk corrections based on mean shift algorithm
Method and its application.
Background technique
After digital pcr technology refers to that DNA or RNA sample is diluted, independent micro- reaction member is assigned to, specific base is passed through
Because the fluorescence probe of fragment label is marked, molecular template PCR amplification is carried out, is sentenced after the completion of amplification according to fluorescence signal power
Break micro- reaction member positive or negative technology.
The fluorescence probe of specific gene fragment label can generate corresponding fluorescence letter after with molecular template chain combination
Number.Different fluorescence all correspond to a specific band, and the sensor for detecting the wave band can show maximum brightness value.The light of fluorescence
Spectrum is continuous, while different fluorescence, there are part intersection, the fluorescence intensity detected on the same wave band may be by several glimmering
Optical superposition, here it is fluorescence crosstalk phenomenons.As shown in Figure 1.
In order to realize that spectra overlap corrects, fluorescence intensity space is switched to dye strength space, and process is as follows: assuming that
In four color fluorescence signals, I1(t), I2(t), I3(t) and I4(t) 4 kinds of fluorescence intensity levels of time t detection, ROX (t), FAM are indicated
(t), VIC (t), CY5 (t) indicate 4 kinds of dye strength values.The conversion formula in dye strength space to fluorescence intensity space is as follows:
In formula, WijIndicate fluorescence intensity accounting of i-th kind of dyestuff under j-th of channel, b1(t), b2(t), B3(t), b4
(t) noise signal in 4 channels is indicated.In above-mentioned mathematical model, S indicates [I1(t), I2(t), I3(t), I4(t)]T, i.e. fluorescence
Signal；F indicates [ROX (t), FAM (t), VIC (t), CY5 (t)]T, i.e. dye signal；W is spectra overlap matrix；Formula indicates
For S=WF.According to fluorescence intensity signals S, dye strength signal is F=W-1S realizes fluorescence crosstalk correction.
Calculating fluorescence crosstalk correction matrix W method at present has peak method, slope iterative method, K-mean cluster etc..
Peak method is manually to select a time point, and the light intensity of n wave band of this time point is inserted correction matrix
Respective column, overlapping comparison is small between signal-to-noise ratio that the wave crest of selection will have, wave crest.However peak method has the disadvantage that
First is that manually selecting wave crest produces human error, second is that sample limits to, all data sets cannot be covered.
K-mean cluster requires fluorescence intensity level to form n cluster in n-dimensional space, but if fluorescence intensity data is interfered
Degree is high or influence of noise forms the cluster for being greater than n, and K-mean cluster may calculate the cluster centre to make mistake, to influence
Correction matrix data.
Slope iterative method is optional two channels as X-axis and Y-axis, and fluorescence intensity level is as coordinate value.Data point is general
It is distributed near X-axis and Y-axis in first quartile, then periodic sampling, the slope of digital simulation straight line inserts corresponding W
Matrix.But if the influence of concentration or noise spot is compared in the distribution of data, after periodic sampling, fitting a straight line, which generates, is greater than 1
Slope, leverage the authenticity of W data.
Summary of the invention
The invention discloses a kind of fluorescence cross-talk correction method based on mean shift algorithm, correction side disclosed by the invention
Method can reduce influence of the noise to fluorescence intensity and can be accurately fast to avoid the human error that introduces when sampling fluorescence intensity level
The element value of correction matrix is calculated fastly, to obtain revised fluorescence data.
Concrete scheme of the invention is as follows:
One aspect of the present invention discloses a kind of fluorescence cross-talk correction method based on mean shift algorithm, including following step
It is rapid:
S1: multichannel fluorescence intensity data pretreatment；
S2: different channel cluster centres are calculated based on mean shift clustering；
S3: calculating slope, fills correction matrix, obtains revised fluorescence data.
It should be understood that the present invention is not limited to above-mentioned steps, the step of can also including other, such as before step S1,
Between step S1 and S2, between step S2 and S3, between step S3 and S4, between step S4 and S5, after step S5, also include
Other additional steps, without beyond the scope of the present invention.
Preferably, the step S1 includes:
S11: under stable imaging system, the chip picture in different channels is shot；
S12: the fluorescent value of micro- reaction member in picture is obtained；
S13: the chip type in channel and picture that selection needs to correct；
S14: fluorescence data pretreatment.
It should be understood that before above-mentioned steps, between, later, include also other additional steps, and without departing from the present invention
Protection scope.
It is furthermore preferred that the S12 includes:
S121: the center of micro- reaction member is positioned；
S122: the mean value of the fluorescent value of micro- reaction member is calculated.
Further, the formula of the mean value of the fluorescent value of micro- reaction member is calculated in S122 are as follows:Wherein, f (x, y) indicates pixel value at center (x, y), Vn(x, y) indicates the
The fluorescent value of n micro- reaction members, n ∈ [1, N], N indicate reaction member number.
It should be understood that before above-mentioned steps, between, later, include also other additional steps, and without departing from the present invention
Protection scope.
Preferably, the S14 includes:
S141: an optional channel fluorescence data value calculatesWherein, Bn(x, y) table
Show the background pixel value of n-th of micro- reaction member,Indicate the new fluorescence data of n-th of micro- reaction member；
S142: maximum fluorescent value in selector channelWith minimum fluorescent valueThe initial threshold is enabled to be
S143: setting data set A and B, wherein A indicates that pixel value is more than or equal to T0Data acquisition system, data amount check M1；B
Indicate that data amount check is less than T0Data acquisition system, data amount check M2；
The gray average G of S144: set of computations A and set BAAnd GB:
S145: new threshold value is calculated
S146: compare front and back iteration T twice, if iteration T is different twice, repeatedly step S143, S144, S145, no
It then carries out in next step；
S147: using T as the datum mark of channel data；To all data point summations greater than T, mean value is obtained Wherein MCPixel value is more than or equal to the data amount check of T；
S148: all data under normalization channel:
WhereinFluorescence intensity level after indicating normalization；
S149: S141-S148 is repeated, the channel data of other selections is normalized.
It should be understood that before above-mentioned steps, between, later, include also other additional steps, and without departing from the present invention
Protection scope.
Preferably, the step S2 includes:
S21: optional normalized two channel Cs H1 and CH2 construct fluorescence intensity data two-dimensional spatial distribution P (Vch1,
Vch2), wherein Vch1And Vch2Respectively indicate the fluorescence intensity level of channel C H1 and channel C H2；
S22: the data point that random selection one is not labeledCentered on point, wherein [1, N] c ∈；
S23: range points are calculatedPoint less than BWNote pointCollection is combined into U,
InWherein BW=0.5, j indicate different in set U
Point；The all the points being arranged in set U are class C, these statistics numbers put are added 1 respectively；
S24: with pointCentered on, it calculatesThe vector of each element into set U, by these
Addition of vectors obtains motion-vector Shift (Vch1, Vch2), updating center position isWherein Wherein j indicates difference in set U；
S25: S23-S24 is repeated, until point Pnew(Vch1, Vch2) and pointDistance less than 0.05, remember at this time
Record point Pnew(Vch1, Vch2), all coordinate points, which are encountered, in this iterative process belongs to class C；
S26: if when convergence current class C Pnew(Vch1, Vch2) have existed at a distance from class center with other less than threshold value
0.1, that merges the two classes, otherwise, using c as new cluster；
S27: S22-S26 is repeated until all labeled access of all points；
S28: statistics each pair of point answers the access frequency of class, that maximum class of access frequency is taken, as current point set
Affiliated class,
S29: all cluster centres that S21-S28 finds any two channel are repeated.
It should be understood that before above-mentioned steps, between, later, include also other additional steps, and without departing from the present invention
Protection scope.
Preferably, the step S3 includes:
S31: slope is calculated；
S32: all slope values take negative, insert in corresponding crosstalk inverse matrix, while the master of crosstalk inverse matrix is diagonal
Element on line is assigned a value of 1；
S33: the inverse matrix of generation is brought into formula F=W-1S obtains revised fluorescence data, wherein W indicates spectrum
Crosstalk signal, S refer to fluorescence intensity signals, and F refers to dye strength signal.
It should be understood that before above-mentioned steps, between, later, include also other additional steps, and without departing from the present invention
Protection scope.
It is furthermore preferred that filtering out cluster centre according to fluorescence intensity level, cluster number and cluster position, then in S31
Calculate slope.
Specifically, in one embodiment of the invention, the S31 includes one in S311 step or S312 step
It is a:
S311: if port number, less than 3, correction matrix master is diagonally all assigned a value of 1, and other positions are assigned a value of 0；
S312: if port number is more than or equal to 3, the calculating of slope is as follows:
S3121: optional two channel Cs H1, CH2 are denoted as X passage and the channel Y, cluster number that copy step S2 is obtained, in
Heart position；
S3122: if cluster number is less than 3, corresponding correction matrix is assigned a value of 0, otherwise carries out following step；
S3123: if cluster number is equal to 3, each cluster centre CiTransverse and longitudinal coordinate value is added WithIndicate the i-th class CiCorresponding transverse and longitudinal coordinate value, CH1 and CH2 indicate corresponding channel；Choosing
It selects and the smallest cluster Cmin, other two cluster centres are denoted as CAAnd CB, then slope
If LI< LII, thenOtherwiseLxAnd LyIndicate channel C H1 and
The corresponding slope of CH2；
S3124: if cluster number is greater than 3, the mean value of all cluster centres is calculated first: Wherein n indicates cluster centre sum.
S3125: calculating slope, and step includes:
Step A1, it screensCluster centre:
(1) if the cluster centre of screening is not present, Lx=0 and Ly=0；
(2) if the cluster centre number of screening is 1, this cluster centre is corresponded to the tables of data in the channel CH1 and CH2
It is shown asWith
(3) if the cluster centre number of screening is greater than 1, by coordinate corresponding to the smallest class of the sum of transverse and longitudinal coordinate value
Value assignsWith
Step A2, it screensCluster centre:
(1) if the cluster centre number of screening is less than 2, Lx=0 and Ly=0；
(2) if the cluster centre number of screening is equal to 2, two cluster centres are denoted as CAAnd CB, then slope If LI< LII, then Lx=LI,Otherwise Lx=LII,
LxAnd LyIndicate the corresponding slope of channel C H1 and CH2；
(3) if the cluster centre number of screening is greater than 2, the corresponding coordinate of the maximum class of abscissa value is denoted as CA,
The corresponding coordinate of the maximum class of ordinate value is denoted as CB, then slopeIf LI< LII, then Lx
=LI,Otherwise Lx=LII,LxAnd LyIndicate the corresponding slope of channel C H1 and CH2；
S3126: repeating step S3121-S3125, calculates the slope value in other channels two-by-two.
It should be understood that before above-mentioned steps, between, later, include also other additional steps, and without departing from the present invention
Protection scope.
On the basis of common knowledge of the art, above-mentioned each optimum condition, can any combination, and without departing from structure of the invention
Think of and protection scope.
The second aspect of the present invention discloses a kind of digital pcr system, adopts with the aforedescribed process, realizes fluorescence crosstalk signal
Correction.
Third aspect of the present invention discloses the sharp application in machine learning techniques field with the aforedescribed process.
The present invention has following remarkable advantage and effect compared with the existing technology:
The uniform drift that the present invention uses clusters bearing calibration, is in the case where being involved in calculating based on all data sets
It completes, the distribution of cluster and shape do not influence the calculating of slope, and largely avoided and occurred using means such as peak methods
The case where selecting single data and sample to limit to.
Data preprocessing phase of the invention uses the method threshold value T of iterative cycles, calculates all fluorescent values greater than T
Data, and calculate their mean value, last renormalization data, final threshold value T is to consider data comprehensively by feminine gender
It is formed with the positive, more accurately.
When average drifting updates cluster centre, formula has been used
Prevent cluster process from falling into local optimum or endless loop.
When calculating slope, method disclosed by the invention is sieved according to situations such as fluorescence intensity level, cluster number, cluster position
True cluster centre is selected, false cluster centre is excluded and participates in calculating, data have more authenticity.
Detailed description of the invention
Fig. 1 is four color fluorescence crosstalk schematic diagrames；
Fig. 2 is technical solution flow chart in the embodiment of the present invention；
Fig. 3 is average drifting schematic diagram in the embodiment of the present invention；
Fig. 4 is fluorescence distribution figure before correcting in the embodiment of the present invention；
Fig. 5 is fluorescence distribution figure after correcting in the embodiment of the present invention.
Specific embodiment
The present invention is further illustrated below by the mode of embodiment, but does not therefore limit the present invention to the reality
It applies among a range.
Method key step disclosed by the invention includes:
S1: multichannel fluorescence intensity data pretreatment；
S2: different channel cluster centres are calculated based on mean shift clustering；
S3: calculating slope, fills correction matrix, obtains revised fluorescence data.
Term is explained in the present invention:
Mean shift algorithm: mean shift clustering is the algorithm based on sliding window, it attempts to find the intensive of data point
Region.This is the algorithm based on mass center, it means that its target be position each group/class central point, by will in
The candidate point of heart point is updated to the mean value put in sliding window to complete.Then, post-processing stages to these candidate windows into
Row filtering with eliminate it is approximate repeat, form final center point set and its corresponding group.
Referring to fig. 2, technical solution of the present invention basic flow chart, including step are shown are as follows:
Data are read in first, then judge port number, if port number is less than triple channel, are calculated slope, are then calculated correction
Matrix obtains revised fluorescence data.If port number is more than or equal to triple channel, solid-state or liquid chip are first selected
Picture then carries out data prediction, then calculates positioning/non-locating channel cluster centre, calculates tiltedly after screening cluster centre
Rate then calculates correction matrix and obtains revised fluorescence data.
Specifically, disclosing a kind of fluorescence crosstalk school based on mean shift algorithm in an embodiment of the invention
Correction method, comprising the following steps:
S1: multichannel fluorescence intensity data pretreatment；
S2: different channel cluster centres are calculated based on mean shift clustering；
S3: calculating slope, fills correction matrix, obtains revised fluorescence data.
In one embodiment of the present invention, the step S1 includes:
S11: under stable imaging system, the chip picture in different channels is shot；
S12: the fluorescent value of micro- reaction member in picture is obtained；
S13: the chip type in channel and picture that selection needs to correct；
S14: fluorescence data pretreatment.
It is furthermore preferred that the S12 includes:
S121: the center of micro- reaction member is positioned；
S122: the mean value of the fluorescent value of micro- reaction member is calculated.
In one embodiment of the present invention, in S122, the formula of the mean value of the fluorescent value of micro- reaction member is calculated are as follows:Wherein, f (x, y) indicates pixel value at center (x, y), Vn(x, y) indicates the
The fluorescent value of n micro- reaction members, n ∈ [1, N], N indicate reaction member number.
In one embodiment of the present invention, the S14 includes:
S141: an optional channel fluorescence data value calculatesWherein, Bn(x, y) table
Show the background pixel value of n-th of micro- reaction member,Indicate the new fluorescence data of n-th of micro- reaction member；
S142: maximum fluorescent value in selector channelWith minimum fluorescent valueThe initial threshold is enabled to be
S143: setting data set A and B, wherein A indicates that pixel value is more than or equal to T0Data acquisition system, data amount check M1；B
Indicate that data amount check is less than T0Data acquisition system, data amount check M2；
The gray average G of S144: set of computations A and set BAAnd GB:
S145: new threshold value is calculated
S146: compare front and back iteration T twice, if iteration T is different twice, repeatedly step S143, S144, S145, no
It then carries out in next step；
S147: using T as the datum mark of channel data；To all data point summations greater than T, mean value is obtained Wherein MCPixel value is more than or equal to the data amount check of T；
S148: all data under normalization channel:
WhereinFluorescence intensity level after indicating normalization；
S149: S141-S148 is repeated, the channel data of other selections is normalized.
In one embodiment of the present invention, the step S2 includes:
S21: optional normalized two channel Cs H1 and CH2 construct fluorescence intensity data two-dimensional spatial distribution P (Vch1,
Vch2), wherein Vch1And Vch2Respectively indicate the fluorescence intensity level of channel C H1 and channel C H2；
S22: the data point that random selection one is not labeledCentered on point, wherein [1, N] c ∈；
S23: range points are calculatedPoint less than BWNote pointCollection is combined into U,
InWherein BW=0.5, j indicate different in set U
Point；The all the points being arranged in set U are class C, these statistics numbers put are added 1 respectively；
S24: with pointCentered on, it calculatesThe vector of each element into set U, by these
Addition of vectors obtains motion-vector Shift (Vch1, Vch2), updating center position isWherein Wherein j indicates difference in set U；
S25: S23-S24 is repeated, until point Pnew(Vch1, Vch2) and pointDistance less than 0.05, remember at this time
Record point Pnew(Vch1, Vch2), all coordinate points, which are encountered, in this iterative process belongs to class C；
S26: if when convergence current class C Pnew(Vch1, Vch2) have existed at a distance from class center with other less than threshold value
0.1, that merges the two classes, otherwise, using c as new cluster；
S27: S22-S26 is repeated until all labeled access of all points；
S28: statistics each pair of point answers the access frequency of class, that maximum class of access frequency is taken, as current point set
Affiliated class,
S29: all cluster centres that S21-S28 finds any two channel are repeated.
Detailed process is as shown in figure 3, center is randomly selected data pointEffective coverage refers to collection
The all the points in U are closed, each point arrivesDistance be less than BW, solid line indicate black arrow indicate motion-vector
Shift(Vch1, Vch2)。
In one embodiment of the present invention, the step S3 includes:
S31: slope is calculated；
S32: all slope values take negative, insert in corresponding crosstalk inverse matrix, while the master of crosstalk inverse matrix is diagonal
Element on line is assigned a value of 1.Table 1 is the inverse matrix for having selected the fluorescence data in one group of 3 channel to be calculated: the elements in a main diagonal
It is 1, non-the elements in a main diagonal takes negative obtain by the element of crosstalk matrix corresponding position.
Inverse matrix is corrected in 1 crosstalk of table
| Channel | ROX | VIC | FAM |
| ROX | 1 | -0.398 | -0.149 |
| VIC | 0 | 1 | -0.337 |
| FAM | 0 | -0.0522 | 1 |
S33: the inverse matrix of generation is brought into formula F=W-1S obtains revised fluorescence data, wherein W indicates spectrum
Crosstalk signal, S refer to fluorescence intensity signals, and F refers to dye strength signal.The display correction front and back channel VIC Fig. 4 and Fig. 5 and the channel FAM
Fluorescence Distribution value, fluorescence Distribution value is more gathered after correction, it is easier to divide corresponding channel the positive and negative point.
In one embodiment of the present invention, in S31, according to fluorescence intensity level, cluster number and the screening of cluster position
Then cluster centre out calculates slope.
Specifically, in one embodiment of the invention, the S31 includes one in S311 step or S312 step
It is a:
S311: if port number, less than 3, correction matrix master is diagonally all assigned a value of 1, and other positions are assigned a value of 0；
S312: if port number is more than or equal to 3, the calculating of slope is as follows:
S3121: optional two channel Cs H1, CH2 are denoted as X passage and the channel Y, cluster number that copy step S2 is obtained, in
Heart position；
S3122: if cluster number is less than 3, corresponding correction matrix is assigned a value of 0, otherwise carries out following step；
S3123: if cluster number is equal to 3, each cluster centre CiTransverse and longitudinal coordinate value is added WithIndicate the i-th class CiCorresponding transverse and longitudinal coordinate value, CH1 and CH2 indicate corresponding channel；Choosing
It selects and the smallest cluster Cmin, other two cluster centres are denoted as CAAnd CB, then slope
If LI< LII, then Lx=LI,Otherwise Lx=LII,LxAnd LyIndicate channel C H1 and CH2
Corresponding slope；
S3124: if cluster number is greater than 3, the mean value of all cluster centres is calculated first: Wherein n indicates cluster centre sum.
S3125: calculating slope, and step includes:
Step A1, it screensCluster centre:
(1) if the cluster centre of screening is not present, Lx=0 and Ly=0；
(2) if the cluster centre number of screening is 1, this cluster centre is corresponded to the tables of data in the channel CH1 and CH2
It is shown asWith
(3) if the cluster centre number of screening is greater than 1, by coordinate corresponding to the smallest class of the sum of transverse and longitudinal coordinate value
Value assignsWith
Step A2, it screensCluster centre:
(1) if the cluster centre number of screening is less than 2, Lx=0 and Ly=0；
(2) if the cluster centre number of screening is equal to 2, two cluster centres are denoted as CAAnd CB, then slope If LI< LII, then Lx=LI,Otherwise Lx=LII,
LxAnd LyIndicate the corresponding slope of channel C H1 and CH2；
(3) if the cluster centre number of screening is greater than 2, the corresponding coordinate of the maximum class of abscissa value is denoted as CA,
The corresponding coordinate of the maximum class of ordinate value is denoted as CB, then slopeIf LI< LII, then Lx
=LI,Otherwise Lx=LII,LxAnd LyIndicate the corresponding slope of channel C H1 and CH2；
S3126: repeating step S3121-S3125, calculates the slope value in other channels two-by-two.
The above embodiment is a preferred embodiment of the present invention, but embodiments of the present invention are not by above-described embodiment
Limitation, other any changes, modifications, substitutions, combinations, simplifications made without departing from the spirit and principles of the present invention,
It should be equivalent substitute mode, be included within the scope of the present invention.
Claims (10)
1. a kind of fluorescence cross-talk correction method based on mean shift algorithm, which comprises the following steps:
S1: multichannel fluorescence intensity data pretreatment；
S2: different channel cluster centres are calculated based on mean shift clustering；
S3: calculating slope, fills correction matrix, obtains revised fluorescence data.
2. the method according to claim 1, wherein the step S1 includes:
S11: under stable imaging system, the chip picture in different channels is shot；
S12: the fluorescent value of micro- reaction member in picture is obtained；
S13: the chip type in channel and picture that selection needs to correct；
S14: fluorescence data pretreatment.
3. according to the method described in claim 2, it is characterized in that, the S12 includes:
S121: the center of micro- reaction member is positioned；
S122: the mean value of the fluorescent value of micro- reaction member is calculated.
4. according to the method described in claim 3, it is characterized in that, calculating the mean value of the fluorescent value of micro- reaction member in S122
Formula are as follows:Wherein, f (x, y) indicates pixel value at center (x, y),
Vn(x, y) indicates that the fluorescent value of n-th of micro- reaction member, n ∈ [1, N], N indicate reaction member number.
5. according to the method described in claim 2, it is characterized in that, the S14 includes:
S141: an optional channel fluorescence data value calculates
Wherein, Bn(x, y) indicates the background pixel value of n-th of micro- reaction member,Indicate that n-th of micro- reaction member is new
Fluorescence data；
S142: maximum fluorescent value in selector channelWith minimum fluorescent valueThe initial threshold is enabled to be
S143: setting data set A and B, wherein A indicates that pixel value is more than or equal to T0Data acquisition system, data amount check M1；B is indicated
Data amount check is less than T0Data acquisition system, data amount check M2；
The gray average G of S144: set of computations A and set BAAnd GB:
S145: new threshold value is calculated
S146: relatively front and back iteration T twice, if iteration T is different twice, repeatedly step S143, S144, S145, otherwise into
Row is in next step；
S147: using T as the datum mark of channel data；To all data point summations greater than T, mean value is obtained Wherein MCPixel value is more than or equal to the data amount check of T；
S148: all data under normalization channel:
WhereinFluorescence intensity level after indicating normalization；
S149: S141-S148 is repeated, the channel data of other selections is normalized.
6. the method according to claim 1, wherein the step S2 includes:
S21: optional normalized two channel Cs H1 and CH2 construct fluorescence intensity data two-dimensional spatial distribution P (Vch1,
Vch2), wherein Vch1And Vch2Respectively indicate the fluorescence intensity level of channel C H1 and channel C H2；
S22: the data point that random selection one is not labeledCentered on point, wherein [1, N] c ∈；
S23: range points are calculatedPoint less than BWNote pointCollection is combined into U,
InWherein BW=0.5, j indicate set U
Middle difference；The all the points being arranged in set U are class C, these statistics numbers put are added 1 respectively；
S24: with pointCentered on, it calculatesThe vector of each element into set U, by these to
Amount is added, and obtains motion-vector Shift (Vch1, Vch2), updating center position is Wherein Wherein j is indicated in set U not
Same point；
S25: S23-S24 is repeated, until point Pnew(Vch1, Vch2) and pointDistance less than 0.05, record at this time
Point Pnew(Vch1, Vch2), all coordinate points, which are encountered, in this iterative process belongs to class C；
S26: if when convergence current class C Pnew(Vch1, Vch2) with it is other have existed at a distance from class center be less than threshold value 0.1,
That merges the two classes, otherwise, using c as new cluster；
S27: S22-S26 is repeated until all labeled access of all points；
S28: statistics each pair of point answers the access frequency of class, that maximum class of access frequency is taken, as belonging to current point set
Class,
S29: all cluster centres that S21-S28 finds any two channel are repeated.
7. the method according to claim 1, wherein the step S3 includes:
S31: slope is calculated；
S32: all slope values take negative, insert in corresponding crosstalk inverse matrix, while on the leading diagonal of crosstalk inverse matrix
Element be assigned a value of 1；
S33: the inverse matrix of generation is brought into formula F=W-1S obtains revised fluorescence data, wherein W indicates spectra overlap
Signal, S refer to fluorescence intensity signals, and F refers to dye strength signal.
8. the method according to the description of claim 7 is characterized in that according to fluorescence intensity level, cluster number and gathering in S31
Class position filters out cluster centre, then calculates slope.
9. a kind of digital pcr system, which is characterized in that using the method as described in any one of claim 1-8, realize glimmering
The correction of optical crosstalk signal.
10. according to right want any one of 1-8 described in application of the method in machine learning techniques field.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910099554.4A CN109859188B (en) | 2019-01-31 | 2019-01-31 | Fluorescence crosstalk correction method based on mean shift algorithm and application thereof |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910099554.4A CN109859188B (en) | 2019-01-31 | 2019-01-31 | Fluorescence crosstalk correction method based on mean shift algorithm and application thereof |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109859188A true CN109859188A (en) | 2019-06-07 |
| CN109859188B CN109859188B (en) | 2021-04-06 |
Family
ID=66897236
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910099554.4A Active CN109859188B (en) | 2019-01-31 | 2019-01-31 | Fluorescence crosstalk correction method based on mean shift algorithm and application thereof |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109859188B (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112345759A (en) * | 2020-11-16 | 2021-02-09 | 三诺生物传感股份有限公司 | Method for detecting fluorescence intensity peak |
| CN112862297A (en) * | 2021-02-02 | 2021-05-28 | 方舟生物安全科技(广州)有限公司 | Real-time monitoring system for pathogenic microorganism Internet of things |
| CN116503471A (en) * | 2023-06-28 | 2023-07-28 | 南开大学 | Single-molecule positioning imaging drift correction method and system for K-time neighbor position cloud picture |
| CN117054338A (en) * | 2023-08-16 | 2023-11-14 | 四川杰莱美科技有限公司 | Parallel light path system based on PCR appearance |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US7402817B2 (en) * | 2004-07-29 | 2008-07-22 | The Research Foundation Of State University Of New York | System and method for cross-talk cancellation in a multilane fluorescence detector |
| CN101688841A (en) * | 2007-06-29 | 2010-03-31 | 霍夫曼-拉罗奇有限公司 | Systems and methods for determining cross-talk coefficients in PCR and other data sets |
| CN101688838A (en) * | 2008-06-11 | 2010-03-31 | 博奥生物有限公司 | A reliable fluorescence correction method for two-color microarray fluorescence system |
| WO2016057923A1 (en) * | 2014-10-09 | 2016-04-14 | Kinetic River Corp. | Particle analysis and sorting apparatus and methods |
| CN106596489A (en) * | 2016-12-19 | 2017-04-26 | 中国科学院苏州生物医学工程技术研究所 | Processing method of fluorescence intensity data in fluorescence droplet detection |
| CN108139325A (en) * | 2015-09-22 | 2018-06-08 | 迈卡提斯公众有限公司 | Crosstalk correction in the multiplexing analysis of biological sample |
-
2019
- 2019-01-31 CN CN201910099554.4A patent/CN109859188B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US7402817B2 (en) * | 2004-07-29 | 2008-07-22 | The Research Foundation Of State University Of New York | System and method for cross-talk cancellation in a multilane fluorescence detector |
| CN101688841A (en) * | 2007-06-29 | 2010-03-31 | 霍夫曼-拉罗奇有限公司 | Systems and methods for determining cross-talk coefficients in PCR and other data sets |
| CN101688838A (en) * | 2008-06-11 | 2010-03-31 | 博奥生物有限公司 | A reliable fluorescence correction method for two-color microarray fluorescence system |
| WO2016057923A1 (en) * | 2014-10-09 | 2016-04-14 | Kinetic River Corp. | Particle analysis and sorting apparatus and methods |
| CN108139325A (en) * | 2015-09-22 | 2018-06-08 | 迈卡提斯公众有限公司 | Crosstalk correction in the multiplexing analysis of biological sample |
| CN106596489A (en) * | 2016-12-19 | 2017-04-26 | 中国科学院苏州生物医学工程技术研究所 | Processing method of fluorescence intensity data in fluorescence droplet detection |
Non-Patent Citations (5)
| Title |
|---|
| CRISTIAN DOMNISORU等: "Cross-talk filtering in four dye fluorescence-based DNA sequencing", 《ELECTROPHORESIS 2000》 * |
| LEI LI等: "An estimate of the cross-talk matrix in fourdye fluorescence-based DNA sequencing", 《ELECTROPHORESIS 1999》 * |
| 林云跃 等: "四色荧光DNA测序中的转换矩阵分析", 《光学仪器》 * |
| 臧留琴 等: "多重定量PCR系统中多色荧光检测和光谱串扰校正方法", 《光学学报》 * |
| 郑华: "DNA分析仪荧光信号采集与处理系统的研究", 《中国博士学位论文全文数据库 信息科技辑（月刊）》 * |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112345759A (en) * | 2020-11-16 | 2021-02-09 | 三诺生物传感股份有限公司 | Method for detecting fluorescence intensity peak |
| CN112345759B (en) * | 2020-11-16 | 2022-07-12 | 三诺生物传感股份有限公司 | Method for detecting fluorescence intensity peak |
| CN112862297A (en) * | 2021-02-02 | 2021-05-28 | 方舟生物安全科技(广州)有限公司 | Real-time monitoring system for pathogenic microorganism Internet of things |
| CN112862297B (en) * | 2021-02-02 | 2022-03-01 | 方舟生物安全科技(广州)有限公司 | Real-time monitoring system for pathogenic microorganism Internet of things |
| CN116503471A (en) * | 2023-06-28 | 2023-07-28 | 南开大学 | Single-molecule positioning imaging drift correction method and system for K-time neighbor position cloud picture |
| CN116503471B (en) * | 2023-06-28 | 2023-08-29 | 南开大学 | Single-molecule positioning imaging drift correction method and system for K-time neighbor position cloud picture |
| CN117054338A (en) * | 2023-08-16 | 2023-11-14 | 四川杰莱美科技有限公司 | Parallel light path system based on PCR appearance |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109859188B (en) | 2021-04-06 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109859188A (en) | A kind of fluorescence cross-talk correction method and its application based on mean shift algorithm | |
| Norton et al. | Detecting hierarchical genome folding with network modularity | |
| Bzikadze et al. | Automated assembly of centromeres from ultra-long error-prone reads | |
| Imakaev et al. | Iterative correction of Hi-C data reveals hallmarks of chromosome organization | |
| Patruno et al. | A review of computational strategies for denoising and imputation of single-cell transcriptomic data | |
| CN101401101B (en) | Methods and systems for identification of DNA patterns through spectral analysis | |
| CN112884712B (en) | Method and related device for classifying defects of display panel | |
| CN109801680B (en) | Tumor metastasis and recurrence prediction method and system based on TCGA database | |
| CN107533593A (en) | Method for identifying unknown microbiota subpopulations from the set with reference to subgroup by mass spectrography | |
| CN115064209B (en) | Malignant cell identification method and system | |
| CN110648721A (en) | Method and device for detecting copy number variation by aiming at exon capture technology | |
| CN110490836A (en) | DPCR microarray images information processing method | |
| CN105849555A (en) | Iterative clustering of sequence reads for error correction | |
| US7136517B2 (en) | Image analysis process for measuring the signal on biochips | |
| EP4238098A1 (en) | Systems and methods for per-cluster intensity correction and base calling | |
| CN104573004B (en) | A kind of double clustering methods of the gene expression data based on double rank genetic computations | |
| CN117708628A (en) | Spatial domain identification method in spatial transcriptome based on map deep learning | |
| Liao et al. | Exploring the connection between ultraviolet/optical variations and radio emission in radio-quiet quasars: clues about the origin of radio emission | |
| CN104537667A (en) | Image cutting method based on direction multi-target genetic clustering algorithm | |
| Deepa et al. | Development of a Fully Automated Image Analysis Method for High Density cDNA and array CGH Microarray Based Genomic Studies | |
| CN116228777B (en) | Concrete stirring uniformity detection method | |
| CN112037854B (en) | Method and system for obtaining tumor methylation marker based on methylation chip data | |
| CN117523559B (en) | Base recognition method and device, gene sequencer and storage medium | |
| CN116434830B (en) | Tumor focus position identification method based on ctDNA multi-site methylation | |
| CN112599251B (en) | Construction method of disease screening model, disease screening model and screening device |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| TR01 | Transfer of patent right | ||
| TR01 | Transfer of patent right |
Effective date of registration: 20211130 Address after: 518101 g1316, Lianxing building, building B, Yihua new village, district 46, Haifu community, Xin'an street, Bao'an District, Shenzhen City, Guangdong Province Patentee after: Pilot medical technology (Shenzhen) Co.,Ltd. Address before: 310000 room 1114, Jin Jun Road, 341 Shui Xiang Road, Jianggan District, Hangzhou, Zhejiang. Patentee before: PILOT GENE TECHNOLOGIES (HANGZHOU) Co.,Ltd. |