CN-121999485-A - Microdroplet partitioning and clustering method, device, medium and system

CN121999485ACN 121999485 ACN121999485 ACN 121999485ACN-121999485-A

Abstract

The present disclosure relates to a droplet dividing and clustering method, device, medium and program, the method is applied to a droplet-based digital nucleic acid amplification quantitative analysis system, the method comprises the steps of reading scattered point data of each droplet in a droplet fluorescent image, including positions and brightness, dividing the droplet fluorescent image into a plurality of grids, sequencing and numbering all grids, forming a subarray by all the droplets in each grid, connecting the subarrays to form a droplet array, mapping the droplet array into a target image with the index of the droplet in the droplet array as an abscissa and the brightness of the droplet as an ordinate, performing image division on the droplet array to obtain a reference droplet, and performing droplet dividing and clustering based on the reference droplet. The method can obviously improve the robustness and reliability during clustering, and particularly solves the problem that the robustness and reliability of threshold segmentation and clustering results are poor under the conditions of different sample concentrations, different yin-yang contrast ratios, unequal light fields and the like.

Inventors

ZHONG LIANGJIAN
ZHAO QINGNAN
BAI KUN
CHEN FAYI
XU XINGHAI

Assignees

迈克医疗电子有限公司

Dates

Publication Date: 20260508
Application Date: 20241122
Priority Date: 20241104

Claims (12)

1. A droplet partitioning and clustering method for a droplet-based digital nucleic acid amplification quantitative analysis system, comprising: reading scattered point data of each droplet in the droplet fluorescence image, including position and brightness; Dividing the droplet fluorescence image into a plurality of grids, sequencing and numbering all grids, forming a subarray by all droplets in each grid, and connecting the subarrays to form a droplet array; Taking the index of the droplet in the droplet array as an abscissa and the brightness of the droplet as an ordinate, mapping the droplet array into a target image with a set size for image segmentation so as to obtain a reference droplet; And performing droplet classification and clustering based on the reference droplets.
2. The method of claim 1, wherein dividing the droplet fluorescence image into a plurality of grids, ordering and numbering all grids, all droplets in each grid forming a subarray, the subarrays being connected to form an array of droplets, comprises: dividing the droplet fluorescence image into a plurality of grids; Sequencing and numbering all grids according to an S-shaped sequence; Determining a grid where each droplet is located according to the position of the droplet; Creating a subarray for each grid, adding each droplet to the subarray of the grid in which it is located; all the subarrays are connected according to the serial numbers of the corresponding grids to form a droplet array.
3. The method of claim 1, wherein mapping the array of droplets into a sized target image for image segmentation to obtain reference droplets comprises: taking the index of the droplet in the droplet array as an abscissa and the brightness of the droplet as an ordinate, mapping the droplet array into a target image with a set size to form a droplet scattering point thermodynamic diagram; And dividing an area with heat meeting preset conditions from the microdroplet scattering point thermodynamic diagram to serve as a reference microdroplet mask, and extracting the reference microdroplet falling into the reference microdroplet mask.
4. A method according to claim 3, wherein mapping the array of droplets into a target image of a set size with the index of the droplets in the array of droplets as the abscissa and the brightness of the droplets as the ordinate, forms a droplet heat map, comprises: And carrying out normalization processing on the mapped target image by adding 1 to the gray value of the mapping position of the droplet on the target image, and converting the normalized target image into an image with a gray range of 0-255 to obtain a micro-droplet scattering point thermodynamic diagram.
5. A method according to claim 3, wherein segmenting the region of the thermodynamic diagram satisfying the predetermined condition from the thermodynamic diagram of the microdroplet as a reference microdroplet mask, extracting the reference microdroplets falling into the reference microdroplet mask, comprises: dividing a target area of the microdroplet scattering point thermodynamic diagram, which meets the preset condition, according to the preset reference microdroplet mask size condition; In the case that the target area is one, determining the target area as a reference droplet mask; In the case of more than one of the target areas, determining the lowermost target area in the upper and lower positional relationship as a reference droplet mask; and extracting the reference droplet falling into the reference droplet mask.
6. The method of claim 1, wherein performing droplet classification and clustering based on the reference droplets comprises: calculating the brightness ratio between each droplet and the reference droplet in the neighborhood of each droplet; generating a corresponding nuclear density curve based on the brightness ratio of the reference droplet and the brightness ratio of the non-reference droplet respectively, and superposing the generated nuclear density curves to obtain a brightness ratio density curve; And extracting peak information in the brightness ratio density curve, and dividing and clustering the droplets in the droplet fluorescence image according to the peak information to obtain a negative droplet cluster and/or a positive droplet cluster.
7. The method of claim 6, wherein calculating the ratio of brightness between each droplet and the reference droplet in its neighborhood comprises: Traversing each droplet in the droplet array, taking the position of the current droplet in the droplet fluorescence image as a center, taking the set multiple of the radius of the current droplet as an initial radius, expanding the radius step by step to iteratively search the reference droplet in the neighborhood of the current droplet, and stopping searching when the number of the searched reference droplets reaches a specified threshold or reaches a set iteration number; after the searching is stopped, calculating the ratio between the brightness of the current droplet and the brightness median value or brightness average value of all searched reference droplets under the condition that the reference droplets are searched, and taking the ratio as the brightness ratio between the current droplet and the reference droplets in the neighborhood of the current droplet, and calculating the ratio between the brightness of the current droplet and the brightness median value or brightness average value of all the reference droplets under the condition that the reference droplets are not searched, and taking the ratio as the brightness ratio between the current droplet and the reference droplets in the neighborhood of the current droplet.
8. The method of claim 5, wherein extracting peak information in the luminance ratio density curve, and dividing and clustering droplets in a droplet fluorescence image according to the peak information, to obtain a negative droplet cluster and/or a positive droplet cluster, comprises: Extracting peak information in the brightness ratio density curve; filtering and combining the extracted peak information; Selecting the first m peaks in the brightness ratio density curve according to the dividing requirement of single PCR detection or multiple PCR detection, determining the brightness ratio corresponding to the valley point based on the peak information of the first m peaks, and determining the dividing line of the negative droplet cluster and the positive droplet cluster based on the brightness ratio; And dividing and clustering the droplets in the droplet fluorescence image by using the dividing line to obtain negative droplet clusters and/or positive droplet clusters.
9. The method as recited in claim 1, further comprising: respectively calculating brightness average values or brightness median values of the negative droplet clusters and the positive droplet clusters, and determining the brightness average values or the brightness median values as inter-cluster brightness threshold values; And respectively taking the distance between the brightness average value or the brightness median value of the negative droplet cluster and the inter-cluster brightness threshold value and the distance between the brightness average value or the brightness median value of the positive droplet cluster and the inter-cluster brightness threshold value as an N sigma threshold value, and correcting the droplets in the negative droplet cluster and the positive droplet cluster which do not meet the corresponding N sigma threshold value.
10. A droplet partitioning and clustering device for use in a droplet-based digital nucleic acid amplification quantitative analysis system, comprising: the reading module is used for reading the scattered point data of each droplet in the droplet fluorescence image, including the position and the brightness; The dividing module is used for dividing the droplet fluorescence image into a plurality of grids, sequencing and numbering all the grids, forming a subarray by all droplets in each grid, and connecting all the subarrays to form a droplet array; the mapping module is used for mapping the droplet array into a target image with a set size for image segmentation by taking the index of the droplet in the droplet array as an abscissa and the brightness of the droplet as an ordinate so as to obtain a reference droplet; and the dividing and clustering module is used for dividing and clustering the droplets based on the reference droplets.
11. A computer readable storage medium, on which a computer program is stored, characterized in that the computer program, when being executed by a processor, implements the steps of the method according to any one of claims 1 to 9.
12. A droplet-based digital nucleic acid amplification quantitative analysis system, comprising: Droplet generation means for generating a plurality of droplets based on the sample; a nucleic acid amplification temperature control device for performing a nucleic acid amplification reaction on the plurality of droplets, and The product signal acquisition device is used for acquiring a product signal after the nucleic acid amplification reaction; The droplet generation device comprises: an open container for storing the generated plurality of droplets and providing a place for performing a nucleic acid amplification reaction; the micro-pipeline is positioned above the open container and is used for loading nucleic acid amplification reaction liquid to be detected, and both ends of the micro-pipeline are provided with openings; Vibration means for driving the microchannel to vibrate reciprocally left and right below the liquid level of the open vessel to continuously generate a plurality of droplets; a fluorescence imaging detection device for taking a fluorescence image of the droplet, and A controller for implementing the steps of the method of any one of claims 1 to 9.

Description

Microdroplet partitioning and clustering method, device, medium and system Cross Reference to Related Applications The present invention claims priority from chinese patent application CN202411560896.9 entitled "droplet classification and clustering method, device, medium and system" filed on month 11 and 4 of 2024, the entire contents of which are incorporated herein by reference. Technical Field The present disclosure relates to the field of image-type digital PCR technology, and in particular, to a droplet partitioning and clustering method, device, medium, and system. Background In the digital PCR (polymerase chain reaction ) technology, the accuracy of clustering and threshold partitioning of the microdroplet data directly affects the accuracy of the result, and related technologies mainly include methods based on K-Means clustering, density clustering, and maximum and minimum intra-class differences. When the method is used for clustering thousands of micro-scattered point data, the problems of low speed, complex parameter selection and adjustment process, difficulty in dividing outlier data and micro-scattered point data when the micro-scattered point data is sparse, poor robustness and reliability when clustering is caused by data differences caused by other factors, and the like exist, and particularly the problems of poor robustness and reliability when clustering is caused by different sample concentrations (extremely low concentration or extremely high concentration), different yin-yang contrasts (especially low contrast), threshold dividing and clustering results when light fields are uneven and the like exist. Disclosure of Invention The disclosure provides a droplet partitioning and clustering method, device, medium and system, which are used for solving the problem that robustness and reliability are poor in clustering due to data difference. In a first aspect, the present disclosure provides a droplet partitioning and clustering method for use in a droplet-based digital nucleic acid amplification quantitative analysis system, the method comprising: reading scattered point data of each droplet in the droplet fluorescence image, including position and brightness; dividing the droplet fluorescence image into a plurality of grids, sequencing and numbering all the grids, forming a subarray by all the droplets in each grid, and connecting the subarrays to form a droplet array; taking the index of the droplet in the droplet array as an abscissa and the brightness of the droplet as an ordinate, mapping the droplet array into a target image with a set size for image segmentation to obtain a reference droplet; droplet classification and clustering is performed based on the reference droplets. In some exemplary embodiments, the droplet fluorescence image is divided into a plurality of grids, all grids are ordered and numbered, all droplets in each grid form a subarray, and the subarrays are connected to form an array of droplets, comprising: Dividing the droplet fluorescence image into a plurality of grids; Sequencing and numbering all grids according to an S-shaped sequence; Determining a grid where each droplet is located according to the position of the droplet; Creating a subarray for each grid, adding each droplet to the subarray of the grid in which it is located; all the subarrays are connected according to the serial numbers of the corresponding grids to form a droplet array. In some exemplary embodiments, mapping an array of droplets into a sized target image for image segmentation to obtain reference droplets includes: taking the index of the droplet in the droplet array as an abscissa and the brightness of the droplet as an ordinate, mapping the droplet array into a target image with a set size to form a droplet scattering point thermodynamic diagram; and dividing an area with heat meeting preset conditions from the microdroplet scattering point thermodynamic diagram to serve as a reference microdroplet mask, and extracting the reference microdroplets falling into the reference microdroplet mask. In some exemplary embodiments, mapping an array of droplets into a target image of a set size with an index of the droplets in the array of droplets as the abscissa and a brightness of the droplets as the ordinate, forms a map of the scattered points of the droplets, comprising: And the gray value of the mapping position of the droplet on the target image is added with 1, the mapped target image is normalized, and the normalized target image is converted into an image with gray range of 0-255, so as to obtain the micro-droplet scattering point thermodynamic diagram. In some exemplary embodiments, segmenting a region with heat satisfying a preset condition from a microdroplet scattering point thermodynamic diagram as a reference microdroplet mask, extracting reference microdroplets falling into the reference microdroplet mask, includes: dividing a target area with the thermodynamic force meeting the