CN1645139A - Method for analysing non-centrifugal urine by image identifying system - Google Patents

Abstract

A method for detecting urinary tangible composition includes starting up sample automatic feeding system; carrying out image-shooting for urine absorbed in counting chamber by microscopic shooting system; carrying out automatic classification, identification and count for obtained image of urinary dregs by image automatic identifying software on computer and outputting results.

Description

Use the not method of centrifugal urine of image identification system analysis

Technical field

What the present invention relates to is the improvement that urine sedimentation detecting method is learned.

Technical background

In arena detects, adopt the centrifuge method microscopy at present, classic method is a slide method always, directly be coated in after urine is centrifugal on the microslide, be placed on the artificial microscopy of microscopically then, the method plate coating thickness can not be controlled, unsoundness, so be difficult for standardization, testing result is difficult for quantitatively.In recent years, can standardization for arena is detected, the tally methods that adopt more, splash in the tally after urine is centrifugal, at the microscopically artificial counting, the grid of fixed size is arranged in the tally, can realize standardization, but it is cumbersome, time-consuming that the method behaviour does, and it is consistent as differences such as centrifugal urine amount, rotating speed, time, reservation arena volumes the result to be difficult for because of operating conditions is different.Streaming detection arena method also is to use more method in recent years in addition, though simple to operate, the not centrifugal urine of use, but the streaming methodology does not meet the necessary quantitative requirement of the urine sample of stipulating in NCCLS and the CCCLS standard, and the accuracy of testing result is still disputable.

Summary of the invention

The invention provides a kind of not method of centrifugal urine of image identification system analysis of using, solving at present in arena detects, method is difficult for that standardization, testing result are difficult for quantitatively, behaviour does the problem that method is cumbersome, time-consuming, the result is not accurate enough.The technical scheme that the present invention takes is:

One, auto injection operation: the 2ul urine sample that will check sucks counting chamber by the liquid flow control system, mixing, this liquid flow control system is connected with counting chamber by pipeline, and counting chamber is the clear glass element of hollow, and the volume of its count block is 2ul;

Two, the urine that sucks counting chamber is carried out image taking by microscope camera system;

Three, the computer picture automatic recognition software is carried out automatic Classification and Identification and counting to the sediment image that photographs:

A. image typing: use the microscopic system that has digital camera, and micro objective is transferred to 400 times, the sediment image scanning in the not centrifugal urine is entered computing machine;

B. the pre-service of image:, improve the base conditioning of effect for the image of input.Mainly comprise: image denoising, illuminance and threshold process are regulated in the picture contrast conversion automatically, image smoothing, the image corrosion, image expansion, edge extracting, image amplifies, reduction operation.For further work is prepared;

C. image segmentation: image segmentation is the feature according to image: the similarity criterion of the gray scale of pixel, color, texture or characteristic set, image pixel is carried out grouping and clustering, the plane of delineation is pressed its feature: the gray scale of pixel, color, texture, the zoning, graphical analysis and data volume senior processing stage such as identification are significantly reduced, keep the information of relevant picture structure feature simultaneously thereafter;

Target in the image or prospect, in their general correspondence image need come out they separation and Extraction in zone specific, that have peculiar property; Image segmentation just is meant image by its feature zoning, and extracts the technology and the process of target; Here feature can be the gray scale, color, texture of pixel etc., and predefined target can corresponding zone, also can corresponding a plurality of zones, adopt many threshold adaptives bunch collection dividing method;

D. cut little image: prospect is surrounded with rectangle frame, and this prospect is the image of cell in the not centrifugal urine, and cuts little picture according to the edge of rectangle frame,

E. feature extraction: for the little image of having cut apart, by extracting its feature: the gray scale of pixel, color, texture, calculate length breadth ratio, roughness, entropy, related coefficient, energy, co-occurrence matrixs etc. are weighted the gained feature, by range of characteristic values red blood cell, leucocyte, cast, crystallization, bacterium are classified; Specific algorithm is as follows:

1. length breadth ratio:

$f_1=(\frac{{\mathrm{\μ}}_{20}-{\mathrm{\μ}}_{02}}{{\mathrm{\μ}}_{20}+{\mathrm{\μ}}_{02}}+1)/2$

2. roughness

At first, size is 2 in the computed image ^k* 2 ^kIn the active window of individual pixel promptly there be the average intensity value of pixel

$A_k(x,y)=\underset{i=x-2^{k-1}}{\overset{x+2^{k-1}-1}{\mathrm{\Σ}}}\underset{j=y-2^{k-1}}{\overset{y+2^{k-1}-1}{\mathrm{\Σ}}}\frac{g(i,j)}{2^{2k}}$

It is poor to calculate mean intensity

E _k，h(x，y)＝|A _k(x+2 ^k-1，y)-A _k(x-2 ^k-1，y)|

E _k，v(x，y)＝|A _k(x，y+2 ^k-1)-A _k(x，y-2 ^k-1)|

Wherein for each pixel, the k value that can make the E value reach maximum (no matter direction) is used for being provided with optimum dimension

S _best(x，y)＝2 ^k；

At last, calculate S in the entire image _BestMean value

$F_{\mathrm{crs}}=\frac{1}{m\×n}\underset{i=1}{\overset{m}{\mathrm{\Σ}}}\underset{j=1}{\overset{n}{\mathrm{\Σ}}}{S}_{\mathrm{best}}(i,j)$

3. entropy:

$H(d,\mathrm{\θ})=-\underset{i,j}{\mathrm{\Σ}}\left\{P(i,j)\right|d,\mathrm{\θ}\}-\log \{P(i,j)|d,\mathrm{\θ}\}$

4. related coefficient:

$C=(d,\mathrm{\θ})=\frac{\underset{i,j}{\mathrm{\Σ}}(i-{\mathrm{\μ}}_x)(j-{\mathrm{\μ}}_y)P(i,j|d,\mathrm{\θ})}{{\mathrm{\σ}}_x{\mathrm{\σ}}_y}$

Wherein:

${\mathrm{\μ}}_x=\underset{i}{\mathrm{\Σ}}i\underset{j}{\mathrm{\Σ}}P(i,j|d,\mathrm{\θ}),{\mathrm{\μ}}_y=\underset{j}{\mathrm{\Σ}}j\underset{i}{\mathrm{\Σ}}P(i,j|d,\mathrm{\θ})$

${\mathrm{\σ}}_x=\underset{i}{\mathrm{\Σ}}{(i-{\mathrm{\μ}}_x)}^2\underset{j}{\mathrm{\Σ}}P(i,j|d,\mathrm{\θ}),{\mathrm{\σ}}_y=\underset{j}{\mathrm{\Σ}}{(j-{\mathrm{\μ}}_y)}^2\underset{i}{\mathrm{\Σ}}P(i,j|d,\mathrm{\θ})$

5. energy:

$E(d,\mathrm{\θ})=\underset{i,j}{\mathrm{\Σ}}{\left\{P(i,j)\right|d,\mathrm{\θ}\}}^2$

6. steadily local:

$L(d,\mathrm{\θ})=\underset{i,j}{\mathrm{\Σ}}\frac{1}{1+{(i-j)}^2}P(i,j|d,\mathrm{\θ})$

F. image recognition and counting: sorted sediment composition is counted;

G. export the result: the output count results, and print examining report.

The liquid flow control system:

Structure is formed: mainly be made up of peristaltic pump, solenoid valve, pressure transducer, liquid level sensor air pump.

Principle of work: the peristaltic pump that controlling liquid flows and the solenoid valve of flow direction, finish the suction sample or the operation of drawing a design by the rotation of peristaltic pump, the logical operation of finishing absorption or getting sample as control electromagnetic valve.

Advantage of the present invention: simple, it is more accurate to count, and meeting the urine sample of stipulating among NCCLS and the CCCLS must be quantitative standards.Provide a kind of shirtsleeve operation method for detecting the urine visible component, remove cumbersome operation from, avoid the cell breakage that causes because of centrifugal, realize automation mechanized operation, and can count on a large scale sample, replace the deviation that sampling observation estimation now causes, make assay more accurate.

We are mixed with (11/ul, 22/ul, 44/ul), use not centrifugal urine counting method and tally method (method that NCCLS recommends) to carry out the counting statistics contrast experiment respectively of high, medium and low 3 concentration standard liquid with red white corpuscle standard particulate.Not centrifugal urine counting method: promptly respectively not centrifugal standard particulate is sucked in the counting chamber by the liquid flow control system, instrument is counted automatically simultaneously.

The tally method: promptly use property arena centrifuge tube dress 10ml standard particulate 1 time, centrifugal 3 minutes of 400G, the supernatant of inclining, the pipe end, keep 0.2ml automatically, is added in the tally with dropper behind the mixing, is placed on the enterprising pedestrian worker's counting of microscope.

The experiment statistics result is as follows:

	Low concentration (individual/ul)	Middle concentration (individual/ul)	High concentration (individual/ul)
				Concentration of standard solution	????11	????22	????55
Not centrifugal counting method	????10.3	????22.4	????52.4
				The tally method	????11	????22.8	????49.4

By T check, two groups of data there was no significant differences (P＜0.05).

Embodiment

One, auto injection operation: the 2ul urine sample that will check sucks counting chamber by the liquid flow control system, mixing, this liquid flow control system is connected with counting chamber by pipeline, and counting chamber is the clear glass element of hollow, and the volume of its count block is 2ul;

Two, the urine that sucks counting chamber is carried out image taking by microscope camera system;

Three, the computer picture automatic recognition software is carried out automatic Classification and Identification and counting to the sediment image that photographs:

A. image typing: use the microscopic system that has digital camera, and micro objective is transferred to 400 times, the sediment image scanning in the not centrifugal urine is entered computing machine;

B. the pre-service of image:, improve the base conditioning of effect for the image of input.Mainly comprise: image denoising, illuminance and threshold process are regulated in the picture contrast conversion automatically, image smoothing, the image corrosion, image expansion, edge extracting, image amplifies, reduction operation.For further work is prepared;

C. image segmentation: image segmentation is the feature according to image: the similarity criterion of the gray scale of pixel, color, texture or characteristic set, image pixel is carried out grouping and clustering, the plane of delineation is pressed its feature: the gray scale of pixel, color, texture, the zoning, graphical analysis and data volume senior processing stage such as identification are significantly reduced, keep the information of relevant picture structure feature simultaneously thereafter;

Target in the image or prospect, in their general correspondence image need come out they separation and Extraction in zone specific, that have peculiar property; Image segmentation just is meant image by its feature zoning, and extracts the technology and the process of target; Here feature can be the gray scale, color, texture of pixel etc., and predefined target can corresponding zone, also can corresponding a plurality of zones, adopt many threshold adaptives bunch collection dividing method;

D. cut little image: prospect is surrounded with rectangle frame, and this prospect is the image of cell in the not centrifugal urine, and cuts little picture according to the edge of rectangle frame,

E. feature extraction: for the little image of having cut apart, by extracting its feature: the gray scale of pixel, color, texture, calculate length breadth ratio, roughness, entropy, related coefficient, energy, co-occurrence matrix etc., the gained feature is weighted, red blood cell, leucocyte, cast, crystallization, bacterium is classified by range of characteristic values; Specific algorithm is as follows:

1. length breadth ratio:

$f_1=(\frac{{\mathrm{\μ}}_{20}-{\mathrm{\μ}}_{02}}{{\mathrm{\μ}}_{20}+{\mathrm{\μ}}_{02}}+1)/2$

2. roughness

At first, size is 2 in the computed image ^k* 2 ^kIn the active window of individual pixel promptly there be the average intensity value of pixel

$A_k(x,y)=\underset{i=x-2^{k-1}}{\overset{x+2^{k-1}-1}{\mathrm{\Σ}}}\underset{j=y-2^{k-1}}{\overset{y+2^{k-1}-1}{\mathrm{\Σ}}}\frac{g(i,j)}{2^{2k}}$

It is poor to calculate mean intensity

E _k，h(x，y)＝|A _k(x+2 ^k-1，y)-A _k(x-2 ^k-1，y)|

E _k，v(x，y)＝|A ^k(x，y+2 ^k-1)-A _k(x，y-2 ^k-1)|

Wherein for each pixel, the k value that can make the E value reach maximum (no matter direction) is used for being provided with optimum dimension

S _best(x，y)＝2 ^k；

At last, calculate S in the entire image _BestMean value

$F_{\mathrm{crs}}=\frac{1}{m\×n}\underset{i=1}{\overset{m}{\mathrm{\Σ}}}\underset{j=1}{\overset{n}{\mathrm{\Σ}}}{S}_{\mathrm{best}}(i,j)$

3. entropy:

$H(d,\mathrm{\θ})=-\underset{i,j}{\mathrm{\Σ}}\left\{P(i,j)\right|d,\mathrm{\θ}\}-\log \{P(i,j)|d,\mathrm{\θ}\}$

4. related coefficient:

$C=(d,\mathrm{\θ})=\frac{\underset{i,j}{\mathrm{\Σ}}(i-{\mathrm{\μ}}_x)(j-{\mathrm{\μ}}_y)P(i,j|d,\mathrm{\θ})}{{\mathrm{\σ}}_x{\mathrm{\σ}}_y}$

Wherein:

${\mathrm{\μ}}_x=\underset{i}{\mathrm{\Σ}}i\underset{j}{\mathrm{\Σ}}P(i,j|d,\mathrm{\θ}),{\mathrm{\μ}}_y=\underset{j}{\mathrm{\Σ}}j\underset{i}{\mathrm{\Σ}}P(i,j|d,\mathrm{\θ})$

${\mathrm{\σ}}_x=\underset{i}{\mathrm{\Σ}}{(i-{\mathrm{\μ}}_x)}^2\underset{j}{\mathrm{\Σ}}P(i,j|d,\mathrm{\θ}),{\mathrm{\σ}}_y=\underset{j}{\mathrm{\Σ}}{(j-{\mathrm{\μ}}_y)}^2\underset{i}{\mathrm{\Σ}}P(i,j|d,\mathrm{\θ})$

5. energy:

$E(d,\mathrm{\θ})=\underset{i,j}{\mathrm{\Σ}}{\left\{P(i,j)\right|d,\mathrm{\θ}\}}^2$

6. steadily local:

$L(d,\mathrm{\θ})=\underset{i,j}{\mathrm{\Σ}}\frac{1}{1+{(i-j)}^2}P(i,j|d,\mathrm{\θ})$

F. image recognition and counting: sorted sediment composition is counted;

G. export the result: the output count results, and print examining report.

Liquid flow control system: mainly form by peristaltic pump, solenoid valve, pressure transducer, liquid level sensor air pump etc.

Principle of work: the peristaltic pump that controlling liquid flows and the solenoid valve of flow direction, finish the suction sample or the operation of drawing a design by the rotation of peristaltic pump, the logical operation of finishing absorption or getting sample as control electromagnetic valve.

Peristaltic pump producer: U.S. Start company, pressure transducer producer: U.S. honeywell company, solenoid valve producer: Japanese TAKSAJO company.

Claims (4)

1, a kind of not method of centrifugal urine of image identification system analysis of using comprises the following steps:

One, auto injection operation: the 2ul urine sample that will check sucks counting chamber by the liquid flow control system, mixing, this liquid flow control system is connected with counting chamber by pipeline, and counting chamber is the clear glass element of hollow, and the volume of its count block is 2ul;

Two, the urine that sucks counting chamber is carried out image taking by microscope camera system;

Three, the computer picture automatic recognition software is carried out automatic Classification and Identification and counting to the sediment image that photographs:

A. image typing: use the microscopic system that has digital camera, and micro objective is transferred to 400 times, the sediment image scanning in the not centrifugal urine is entered computing machine;

B. the pre-service of image:, improve the base conditioning of effect for the image of input.Mainly comprise: image denoising, illuminance and threshold process are regulated in the picture contrast conversion automatically, image smoothing, the image corrosion, image expansion, edge extracting, image amplifies, reduction operation.For further work is prepared;

C. image segmentation: according to the feature of image: the similarity criterion of the gray scale of pixel, color, texture or characteristic set, image pixel is carried out grouping and clustering, the plane of delineation is pressed its feature: the gray scale of pixel, color, texture, the zoning, graphical analysis and data volume senior processing stage such as identification are significantly reduced, keep the information of relevant picture structure feature simultaneously thereafter;

D. cut little image: prospect is surrounded with rectangle frame, and this prospect is the image of cell in the not centrifugal urine, and cuts little picture according to the edge of rectangle frame,

E. feature extraction: for the little image of having cut apart, by extracting its feature: the gray scale of pixel, color, texture, calculate length breadth ratio, roughness, entropy, related coefficient, energy, co-occurrence matrixs etc. are weighted the gained feature, by range of characteristic values red blood cell, leucocyte, cast, crystallization, bacterium are classified;

F. image recognition and counting: sorted sediment composition is counted;

G. export the result: the output count results, and print examining report.

2, a kind of not method of centrifugal urine of image identification system analysis of using according to claim 1, image segmentation in the step 3: just be meant image by its feature zoning, and extract the technology and the process of target; Here feature can be the gray scale, color, texture of pixel etc., and predefined target can corresponding zone, also can corresponding a plurality of zones, adopt many threshold adaptives bunch collection dividing method; Target in the image or prospect, zone specific, that have peculiar property in their general correspondence image.

3, a kind of not method of centrifugal urine of image identification system analysis of using according to claim 1, e. feature extraction specific algorithm is as follows in the step 3:

1. length breadth ratio:

$f_1=(\frac{{\mathrm{\μ}}_{20}-{\mathrm{\μ}}_{02}}{{\mathrm{\μ}}_{20}+{\mathrm{\μ}}_{02}}+1)/2$

2. roughness

At first, size is 2 in the computed image ^k* 2 ^kIn the active window of individual pixel promptly there be the average intensity value of pixel

$A_k(x,y)=\underset{i=x-2^{k-1}}{\overset{x+2^{k-1}-1}{\mathrm{\Σ}}}\underset{j=y-2^{k-1}}{\overset{y+2^{k-1}-1}{\mathrm{\Σ}}}\frac{g(i,j)}{2^{2k}}$

It is poor to calculate mean intensity

E _k，h(x，y)＝|A _k(x+2 ^k-1，y)-A _k(x-2 ^k-1，y)|

E _k，v(x，y)＝|A _k(x，y+2 ^k-1)-A _k(x，y-2 ^k-1)|

Wherein for each pixel, the k value that can make the E value reach maximum (no matter direction) is used for being provided with optimum dimension S _Best(x, y)=2 ^k

At last, calculate S in the entire image _BestMean value

$F_{\mathrm{crs}}=\frac{1}{m\×n}\underset{i=1}{\overset{m}{\mathrm{\Σ}}}\underset{j=1}{\overset{n}{\mathrm{\Σ}}}{S}_{\mathrm{best}}(i,j)$

3. entropy:

$H(d,\mathrm{\θ})=-\underset{i,j}{\mathrm{\Σ}}\left\{P(i,j)\right|d,\mathrm{\θ}\}-\log \{P(i,j)|d,\mathrm{\θ}\}$

4. related coefficient:

$C=(d,\mathrm{\θ})\frac{\underset{i,j}{\mathrm{\Σ}}(i-{\mathrm{\μ}}_x)(j-{\mathrm{\μ}}_y)P(i,j|d,\mathrm{\θ})}{{\mathrm{\σ}}_x{\mathrm{\σ}}_y}$

Wherein: ${\mathrm{\μ}}_x=\underset{i}{\mathrm{\Σ}}i\underset{j}{\mathrm{\Σ}}P(i,j|d,\mathrm{\θ})$ ${\mathrm{\μ}}_y=\underset{j}{\mathrm{\Σ}}j\underset{i}{\mathrm{\Σ}}P(i,j|d,\mathrm{\θ})$

${\mathrm{\σ}}_x=\underset{i}{\mathrm{\Σ}}{(i-{\mathrm{\μ}}_x)}^2\underset{j}{\mathrm{\Σ}}P(i,j|d,\mathrm{\θ})$ ${\mathrm{\σ}}_y=\underset{j}{\mathrm{\Σ}}{(j-{\mathrm{\μ}}_y)}^2\underset{i}{\mathrm{\Σ}}P(i,j|d,\mathrm{\θ})$

5. energy:

$E(d,\mathrm{\θ})=\underset{i,j}{\mathrm{\Σ}}{\left\{P(i,j)\right|d,\mathrm{\θ}\}}^2$

6. steadily local:

$L(d,\mathrm{\θ})=\underset{i,j}{\mathrm{\Σ}}\frac{1}{1+{(i-j)}^2}P(i,j|d,\mathrm{\θ})$

4, a kind of not method of centrifugal urine of image identification system analysis of using according to claim 1, the liquid flow control system comprises: mainly by peristaltic pump, solenoid valve, pressure transducer, liquid level sensor air pump.