A method for continuous automatic search of flow cell focal length for lensless imaging

Abstract

The present invention discloses a method for continuous automatic search of focal length of flow cells for lensless imaging. The method pre-analyzes the characteristics of a focus judgment function curve of a system to determine characteristic parameters. Then, in actual use, a two-step search method is used to select a suitable focus judgment function in each step to continuously search for the optimal focal length of flow cells in microfluidics. When searching for a current cell, the optimal focal length of the previous cell is directly used for the second-step search. The first-step search is repeated by setting appropriate constraints to avoid obtaining a local optimum, thereby improving the accuracy of the automatic search. At the same time, compared with the traditional search method, the calculation load and storage load of the POCT device end are effectively reduced.

Description

Technical Field

[0001] The invention belongs to the technical field of lens-free imaging and relates to a continuous automatic search method for the focal length of a flow cell used for lens-free imaging. Background Art

[0002] In applications such as cell detection and cell culture monitoring, lensless imaging technology has gradually occupied an important position in the research of point-of-care testing (POCT) equipment. Lensless holographic imaging technology restores the amplitude and phase information of the object through a phase iteration algorithm. In this process, it is necessary to accurately measure the distance from the object to the imaging surface, that is, the focal length, which is difficult to accomplish, so most of them use an estimation method to give the focal length. At present, there are some automatic focusing algorithms that judge the focal length through a focus judgment function and use various search strategies to automatically search for the optimal focal length. However, most of these automatic search strategies have a narrow search range, are only effective within a certain range of the focus judgment function, and are easily affected by the background. The distance between the living cells flowing in the microfluidics and the imaging surface is larger than the distance between the cells sandwiched in the glass slide and the imaging surface, so when automatically focusing, the search range is also larger. Existing search strategies, such as exhaustive methods or traversal methods, make the computational load and storage load on the POCT device side too large. Summary of the invention

[0003] The purpose of the present invention is to provide a method for continuous automatic search of flow cell focal length for lensless imaging, which can realize continuous automatic search of the optimal focal length of flow cells in microfluidics and effectively reduce the computational load and storage load of the POCT device end.

[0004] The technical solution adopted by the present invention is a method for continuous automatic search of the focal length of a flow cell for lensless imaging, which specifically comprises the following steps:

[0005] Step 1: After the system parameters are determined, the diffraction images of some cells are randomly collected to analyze the jitter characteristics, single peak characteristics and robustness of the focus judgment function curve under the system parameters, and a focus judgment function f1 with good single peak characteristics and small jitter in the overall trend and a focus judgment function f2 with good single peak characteristics and robustness in a certain focal length area are selected;

[0006] Step 2: In an actual lensless imaging system, when a cell flows by, a focus judgment function f1 with a unique peak in a wide range S and a trend is selected, and a search algorithm is used to search with a step length Δs1, where the step length Δs1 is updated and satisfies |Δs1|≥nΔs, where n is a real number greater than 1. Under this condition, the intermediate optimal focal length searched is z mid , the value of the focus judgment function f1 at this time is f1(z mid );

[0007] Step 3. The best focal length z in step 2 mid Based on the selection range [z mid -w,z mid +w], where w is z mid The width of the left and right extensions, as well as the focus judgment function f2 with good unimodality and robustness in the focal length area, are searched with a step length Δs2, where the step length Δs2 is updated. Under this condition, the optimal focal length searched is z best , the value of the focus judgment function f2 at this time is f2(z best );

[0008] Step 4: Based on the optimal focal length z in step 3 best , restore the image of the cell;

[0009] Step 5: When the next cell flows through, the optimal focal length z of the previous cell best As the intermediate optimal focal length z′ of the current cell mid Repeat step 3 to calculate the optimal focal length z′ of the current cell best At this time, the value of the focus judgment function f2 is f2(z′ best );

[0010] Step 6: Compare the focal length z of the previous cell best The focus judgment function value f2(z best ) and the current cell focal length z′ best The focus judgment function value f2(z′ best ), let h be the limited range of the focus judgment function f2 at the cell focal length, when |f2(z best )-f2(z′ best )|>h, the optimal focal length z′ of the current cell is considered best If it is false, repeat steps 2 and 3 to automatically search for the best focal length of the current cell again. best )-f2(z′ best )|≤h, the optimal focal length z′ of the current cell is considered best If true, repeat step 4 to restore the diffraction image of the current cell and set the optimal focal length z′ of the current cell best The best focal length z″ for the next cell mid ;

[0011] Step 7: Repeat steps 3, 4, 5, and 6 to continuously and automatically search for focus and perform diffraction recovery on the diffraction image of the cell flowing through the microfluidic system.

[0012] The present invention is also characterized in that:

[0013] System parameters include light source wavelength, light source intensity, image sensor, and microfluidic scale.

[0014] The focus judgment function in step 1 includes: grayscale focus judgment functions based on grayscale distribution variance, sobel grayscale gradient or sobel grayscale gradient variance, Laplace grayscale gradient or Laplace grayscale gradient variance, and high-frequency focus judgment functions based on Fourier transform, wavelet transform, discrete cosine transform, etc.

[0015] The jitter characteristics of the focus judgment function curve in step 1 are specifically that when the distance z between the imaging surface and the cell is within a certain wide range S, after restoring the holographic image or diffraction image with a step length ΔS, the curve formed by connecting the values ​​of a certain focus judgment function f is calculated as the focus judgment function curve, and within a certain range [s, s+Δs], it satisfies The focus judgment function curve can still reflect an overall upward or downward trend within this range, and the minimum value of Δs describes the jitter characteristics of the focus judgment function in the system.

[0016] The unimodal characteristic refers to the monotonicity and uniqueness of the focus judgment function curve at the peak value, and the robustness refers to the steepness of the focus judgment function curve at the peak value. The steeper the peak value, the better the robustness of the curve at the peak value.

[0017] The search algorithm in step 2 and step 3 is a "blind hill climbing" search algorithm or a gradient descent search algorithm.

[0018] The restoration method in step 4 is: a restoration method based on deconvolution, a restoration method based on TIE, a restoration method based on angular spectrum theory, a restoration method based on Fresnel diffraction, etc.

[0019] The beneficial effects of the present invention are:

[0020] The present invention discloses a method for continuous automatic search of the focal length of a flow cell for lensless imaging. The method pre-analyzes the characteristics of a system focus judgment function curve to determine characteristic parameters, and uses a two-step search method to select a suitable focus judgment function in each step to continuously search for the optimal focal length of the flow cell in the microfluidic system. When searching for the current cell, the optimal focal length of the previous cell is directly used for the second-step search. By setting appropriate constraints for the first-step search, the accuracy of the automatic search is improved. At the same time, compared with the traditional search method, the calculation load and storage load on the POCT device end are effectively reduced. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] FIG1 is a flow chart of a method for continuous automatic search of the focal length of a flow cell for lensless imaging according to the present invention, wherein FIG1(a) is a preliminary analysis of system characteristics, and FIG1(b) is a flow chart of the automatic search method;

[0022] Fig. 2 is a diffraction image of a cell and a plurality of focus judgment function curves of a flow cell focus continuous automatic search method for lensless imaging of the present invention, wherein Fig. 2(a) is a diffraction image of the cell, Fig. 2(b) is a focus judgment function curve based on cosine score, Fig. 2(c) is a focus judgment function curve based on grayscale variance, Fig. 2(d) is a focus judgment function curve based on Fourier transform, Fig. 2(e) is a focus judgment function curve based on wavelet transform, and Fig. 2(f) is a focus judgment function curve based on contrast;

[0023] Figure 3 It is a schematic diagram of the search process of the first step of a method for continuous automatic search of the focal length of a flow cell for lensless imaging of the present invention;

[0024] Figure 4 It is a schematic diagram of the search process of the second step of a method for continuous automatic search of the focal length of a flow cell for lensless imaging of the present invention;

[0025] Figure 5 It is a restored image of the cell diffraction image in FIG2(a) of a method for continuous automatic search of the focal length of a flow cell for lensless imaging of the present invention;

[0026] FIG6 is a schematic diagram of the second step search process of a cell next to the cell in FIG2(a) of the method for continuous automatic search of the focal length of a flow cell for lensless imaging of the present invention, wherein FIG6(a) is a schematic diagram of the second step automatic search of the focal length of the current cell being judged as false, and FIG6(a) is a schematic diagram of the second step automatic search of the focal length of the current cell being judged as true; DETAILED DESCRIPTION

[0027] The present invention is described in detail below with reference to the accompanying drawings and specific embodiments.

[0028] The present invention provides a method for continuous automatic search of the focal length of a flow cell for lensless imaging. The method can be used for restoring a holographic image or a diffraction image by a lensless area array image sensor imaging system or a line array scanning imaging system based on microfluidics, as shown in FIG1 , and specifically comprises the following steps:

[0029] Step 1, after the system parameters such as light source wavelength, light source intensity, image sensor, and microfluidic scale are determined, the diffraction images of some cells are randomly collected to analyze the jitter characteristics, single peak characteristics, and robustness of some commonly used focus judgment function curves under the system parameters, and select a focus judgment function f1 with good single peak characteristics and small jitter in the overall trend, and a focus judgment function f2 with good single peak characteristics and robustness in a certain focal length area;

[0030] Common focus judgment functions may include: grayscale focus judgment functions based on grayscale distribution variance, sobel grayscale gradient or sobel grayscale gradient variance, Laplace grayscale gradient or Laplace grayscale gradient variance, and high-frequency focus judgment functions based on Fourier transform, wavelet transform, discrete cosine transform, etc.;

[0031] The jitter characteristics of the focus judgment function curve are analyzed as follows: when the distance z between the imaging surface and the cell is within a certain wide range S, after restoring the holographic image or diffraction image with a step length ΔS, the curve formed by connecting the values ​​of a certain focus judgment function f is the focus judgment function curve, and within a certain range [s, s+Δs], it satisfies The focus judgment function curve can still reflect the overall upward or downward trend within this range, and the minimum value of Δs describes the jitter characteristics of the focus judgment function in the system;

[0032] The unimodal characteristic refers to the monotonicity and uniqueness of the focus judgment function curve at the peak value, and the robustness refers to the steepness of the focus judgment function curve at the peak value. The steeper the peak value, the better the robustness of the curve at the peak value.

[0033] As shown in Figure 2, it is a diffraction image of a certain cell, and various focus judgment function curves, including focus judgment functions based on wavelet transform, contrast, cosine score, Fourier transform, grayscale variance, etc. It can be found that the focus judgment functions based on grayscale variance and contrast show a single peak in the overall trend and a small jitter, which can be used as the focus judgment function f1. The focus judgment functions based on Fourier transform and wavelet transform have good single peak and robustness in the focal length area, and can be used as the focus judgment function f2.

[0034] Step 2: In an actual lensless imaging system, when a cell flows by, a focus judgment function f1 with a unique peak in a wide range S and a trend is selected, and a search algorithm (a "blind hill climbing" search algorithm or a gradient descent search algorithm) is used to search with a step length Δs1, where the step length Δs1 is updated and satisfies |Δs1|≥nΔs, where n is a real number greater than 1. Under this condition, the intermediate optimal focal length searched is z mid , the value of the focus judgment function f1 at this time is f1(z mid ),

[0035] like Figure 3 As shown, given an initial value z init , and the initial step length Δs1 = 3Δs. After three iterative searches, Δs1 gradually decreases, and the intermediate optimal focal length z in step 2 is determined. mid , and the focus judgment function value f1(z mid );

[0036] Step 3. The best focal length z in step 2 midBased on the selection range [z mid -w,z mid +w], where w is z mid The width of the left and right extensions, as well as the focus judgment function f2 with good unimodality and robustness in the focal length area, are searched with a step length Δs2 (using the "blind hill climbing" search algorithm or the gradient descent search algorithm), where the step length Δs2 is updated. Under this condition, the optimal focal length searched is z best , the value of the focus judgment function f2 at this time is f2(z best );

[0037] like Figure 4 As shown, the search range [z mid -w,z mid +w], where w = Δs, after three searches, the optimal focal length z is found best , and the focus judgment function value f2(z best );

[0038] Step 4: Based on the optimal focal length z in step 3 best , restore the cell image, the restoration method is: based on the deconvolution restoration method or based on the TIE (Transport of Intensity Equation) restoration method or based on the angular spectrum theory restoration method or based on the Fresnel diffraction restoration method, such as Figure 5 , is the restored image of the cell diffraction image in Figure 2(a);

[0039] Step 5: When the next cell flows through, the optimal focal length z of the previous cell best As the intermediate optimal focal length z′ of the current cell mid Repeat step 3 to calculate the optimal focal length z′ of the current cell best At this time, the value of the focus judgment function f2 is f2(z′ best ),

[0040] Step 6: Compare the focal length z of the previous cell best The focus judgment function value f2(z best ) and the current cell focal length z′ best The focus judgment function value f2(z′ best ), let h be the limited range of the focus judgment function f2 at the cell focal length, when |f2(z best )-f2(z′ best )|>h, the optimal focal length z′ of the current cell is considered best is false, repeat steps 2 and 3 to automatically search for the best focal length of the current cell again, as shown in Figure 6(a), which is a schematic diagram of the second step of automatic search for focal length of the current cell being false; when |f2(z best)-f2(z′ best )|≤h, the optimal focal length z′ of the current cell is considered best As shown in Figure 6(b), the second step of automatic search for focal length of the current cell is true. Repeat step 4 to restore the diffraction image of the current cell and set the optimal focal length z′ of the current cell to best The best focal length z″ for the next cell mid ;

[0041] Step 7: Repeat steps 3, 4, 5, and 6 to continuously and automatically search for focus and perform diffraction recovery on the diffraction image of the cell flowing through the microfluidic system.

[0042] From the above content, it can be seen that the method for continuous automatic search of flow cell focal length for lensless imaging of the present invention can realize continuous automatic search of the optimal focal length of flow cells in microfluidics, effectively reducing the computational load and storage load on the POCT device side.

Claims

1. A method for continuous automatic search of flow cell focal length for lensless imaging, comprising the following steps: Step 1: After the system parameters are determined, randomly collect the diffraction images of some cells, analyze the jitter characteristics, single peak characteristics and robustness of the focus judgment function curve under the system parameters, and select a focus judgment function with good single peak characteristics and small jitter in the overall trend. , and a focus judgment function with good unimodality and robustness in a certain focal length region ; The jitter characteristics of the focus judgment function curve analyzed in step 1 are specifically as follows: when the distance z between the imaging surface and the cell is within a certain wide range Within, with a step length , after restoring the holographic image or diffraction image, calculate a focus judgment function The curve formed by connecting the values ​​of is the curve of the focus judgment function. Inside, satisfied , the focus judgment function curve can still reflect the overall upward or downward trend within this range, then The minimum value of describes the jitter characteristics of the focus judgment function in this system; The unimodal characteristic refers to the monotonicity and uniqueness of the focus judgment function curve at the peak value, and the robustness refers to the steepness of the focus judgment function curve at the peak value. The steeper the peak value, the better the robustness of the curve at the peak value. Step 2: In the actual lensless imaging system, when a cell flows by, select The focus judgment function with a unique peak in the trend , using a search algorithm with a step size Search with a step size of There is an update and it satisfies , n is a real number greater than 1, and the intermediate optimal focal length to be searched is , the focus judgment function at this time The value of ; Step 3: The best focal length in step 2 Based on the selection range , where w is The width of the left and right extensions, as well as the single-peaked and robust focus judgment function in the focal length area , with one step length Search with a step size of There is an update, and the best focal length searched at this time is , the focus judgment function at this time The value of ; Step 4: Based on the best focal length in step 3 , restore the image of the cell; Step 5: When the next cell flows through, the best focal length of the previous cell As the middle best focal length of the current cell , repeat step 3 to calculate the optimal focal length of the current cell , the focus judgment function at this time The value of ; Step 6: Compare the focal length of the previous cell Time-dependent focus function value The focal length of the current cell Time-dependent focus function value Let h be the focus judgment function at the cell focal length. When the limit of When the optimal focal length of the current cell is If false, repeat steps 2 and 3 to automatically search for the best focal length of the current cell. When the optimal focal length of the current cell is If true, repeat step 4 to restore the diffraction image of the current cell and set the best focal length of the current cell to The best focal distance for the middle of the next cell ; Step 7: Repeat steps 3, 4, 5, and 6 to continuously and automatically search for focus and perform diffraction recovery on the diffraction image of the cell flowing through the microfluidic system.

2. A method for continuous automatic search of flow cell focal length for lensless imaging according to claim 1, characterized in that: The system parameters include light source wavelength, light source intensity, image sensor, and microfluidic scale.

3. The method for continuous automatic search of flow cell focal length for lensless imaging according to claim 1, characterized in that: The focus judgment function in step 1 includes: grayscale focus judgment functions based on grayscale distribution variance, sobel grayscale gradient or sobel grayscale gradient variance, Laplace grayscale gradient or Laplace grayscale gradient variance, and high-frequency focus judgment functions based on Fourier transform, wavelet transform, discrete cosine transform, etc.

4. The method for continuous automatic search of flow cell focal length for lensless imaging according to claim 1, characterized in that: The search algorithm in step 2 and step 3 is a "blind hill climbing" search algorithm or a gradient descent search algorithm.

5. The method for continuous automatic search of flow cell focal length for lensless imaging according to claim 1, characterized in that: The restoration method in step 4 is: a restoration method based on deconvolution, a restoration method based on TIE, a restoration method based on angular spectrum theory, or a restoration method based on Fresnel diffraction.

Images