CN-122023394-A - Machine learning-based high-throughput data processing method and system for focused micro-pore plate

CN122023394ACN 122023394 ACN122023394 ACN 122023394ACN-122023394-A

Abstract

The invention relates to the technical field of data processing, and provides a machine learning-based high-throughput data processing method and system for a focused micro-pore plate. Then, by combining the high-throughput image data and a plurality of quality detection viewpoints, initial quality inspection discrimination information is generated. This information is then sent to a specially designed microplate quality assurance network for processing. The network can effectively carry out deep analysis and decision on the initial quality inspection discrimination information by learning a large number of microplate image data with labels, and finally outputs an accurate microplate quality inspection viewpoint as a discrimination result. The process realizes the automatic quality evaluation of the micro-pore plate, remarkably improves the working efficiency and the discrimination accuracy, and provides powerful technical support for experimental study in the biomedical field.

Inventors

ZHANG LIANG
GONG LISHA

Assignees

贵州理工学院
深圳市瑾蕙电子科技有限公司

Dates

Publication Date: 20260512
Application Date: 20260410

Claims (10)

1. A machine learning-based high throughput data processing method for a focused microplate, the method being applied to an image data processing system, the method comprising: acquiring a to-be-processed microplate image and a plurality of microplate quality detection viewpoints, wherein the to-be-processed microplate image comprises microplate type to be detected and high-throughput image data; Combining the quality detection viewpoints of the plurality of micro-pore plates and the high-throughput image data of the micro-pore plate image to be processed to generate initial quality inspection discrimination information; Performing microplate quality detection viewpoint decision on the initial quality detection discrimination information through a microplate quality detection discrimination network, and determining a microplate quality detection discrimination result of the to-be-processed microplate image, wherein the microplate quality detection discrimination result is one of the plurality of microplate quality detection viewpoints; The debugging step of the micro-pore plate quality inspection discrimination network comprises the following steps: obtaining X network training samples, wherein X is an integer greater than 0; For each network training sample, acquiring the type of the to-be-detected micro-plate of the current network training sample as a target micro-plate type and acquiring the type of the to-be-detected micro-plate of Y rest network training samples from X-1 network training samples as an anti-micro-plate type, wherein Y is an integer greater than 0 and less than the number of set viewpoints, and the number of set viewpoints is the maximum number of priori quality detection viewpoints in the micro-plate quality inspection discrimination results of micro-plate quality detection viewpoint decision; performing image combination according to the target micro-pore plate type, the anti-micro-pore plate type and the high-throughput image data of each network training sample to generate a quality inspection distinguishing information sample, and obtaining a quality inspection distinguishing information sample set, wherein the target micro-pore plate type is a matching type corresponding to the quality inspection distinguishing information sample, the quality inspection distinguishing information sample set comprises a plurality of quality inspection distinguishing information samples, and each quality inspection distinguishing information sample is correspondingly generated through one network training sample; And debugging the basic quality inspection discrimination network based on the quality inspection discrimination information sample set to obtain the micro-pore plate quality inspection discrimination network.
2. The method of claim 1, wherein the combining the high-throughput image data based on the plurality of microplate quality inspection perspectives and the microplate image to be processed to generate initial quality inspection discriminatory information comprises: Image combination is carried out on the quality detection viewpoints of the plurality of micro-pore plates and the high-flux image data of the micro-pore plate image to be processed, so that joint discrimination information is obtained; If the number of the micro-pore plate quality detection views is smaller than the total number of the micro-pore plate quality detection and discrimination network, training notes are configured to the combined discrimination information based on the comparison result of the total number of the micro-pore plate quality detection views and the number of the micro-pore plate quality detection views, so that initial quality detection discrimination information is obtained.
3. The method of claim 2, wherein image combining the plurality of microplate quality inspection perspectives with the high-throughput image data of the microplate image to be processed to obtain joint discrimination information comprises: Performing image mapping on the quality detection viewpoints of the microplates according to the microplates reference images corresponding to the quality detection viewpoints of the microplates to obtain a plurality of quality inspection mapping image data; And carrying out image combination on the quality inspection mapping image data and the high-flux image data of the to-be-processed micro-pore plate image to obtain joint discrimination information.
4. The method of claim 1, wherein said performing a microplate quality inspection point of view decision on said initial quality inspection discrimination information via said microplate quality inspection discrimination network, determining a microplate quality inspection discrimination result of said microplate image to be processed, comprises: performing microplate quality detection viewpoint decision on the initial quality inspection discrimination information through a microplate quality inspection discrimination network to obtain quality detection viewpoint decision information; If the number of decision sequences of the quality detection viewpoint decision information is larger than the number of viewpoints of the quality detection viewpoints of the plurality of micro-pore plates, extracting the quality detection viewpoint decision information based on the number of viewpoints to obtain decision query characteristics; And acquiring the microplate quality detection viewpoint corresponding to the decision query feature as a microplate quality inspection discrimination result of the microplate image to be processed according to the priorities of the plurality of microplate quality detection viewpoints in the initial quality inspection discrimination information.
5. The method according to claim 1, wherein for each network training sample, obtaining the type of microplate to be measured in the current network training sample as the target microplate type and obtaining the type of microplate to be measured in Y remaining network training samples from the X-1 network training samples as the counter microplate type comprises: any one network training sample in the X network training samples is obtained and used as a current network training sample; in a quantization constraint interval with the prior viewpoint number generated by the basic quality inspection discrimination network as a maximum value, determining an arbitrary value as an countermeasure viewpoint number Y; and based on the number Y of the countermeasure views, obtaining network training samples except Y from the X-1 network training samples as the rest network training samples.
6. The method of claim 5, wherein the generating a quality control discrimination information sample based on the high throughput image data of the target microplate type, the counter microplate type, and the current network training sample comprises: Randomly combining the target micro-pore plate type with the countermeasure micro-pore plate type to obtain joint discrimination information; If the number of the training samples of the target microplate type and the countermeasure microplate type is smaller than the total number of the points of view of the microplate quality inspection discrimination network, configuring training notes to the joint discrimination information based on the comparison result of the total number of the points of view and the number of the training samples; And combining the combined discrimination information with the high-throughput image data of the current network training sample to obtain a quality inspection discrimination information sample.
7. The method of claim 1, wherein the debugging of the basic quality control discrimination network based on the quality control discrimination information sample set, before obtaining a microplate quality control discrimination network, further comprises generating an challenge training image based on image combinations of the target microplate type, the challenge microplate type, and high throughput image data of the current network training sample, and adding the challenge training image to the quality control discrimination information sample set, wherein the challenge microplate type is a matching type corresponding to the challenge training image; The method comprises the steps of debugging a basic quality inspection discrimination network based on the quality inspection discrimination information sample set to obtain a micro-pore plate quality inspection discrimination network, wherein the basic quality inspection discrimination network is debugged through quality inspection discrimination information samples and countermeasure training images in the quality inspection discrimination information sample set to obtain the micro-pore plate quality inspection discrimination network.
8. The method of claim 7, wherein the debugging the basic quality inspection discrimination network through the quality inspection discrimination information samples and the challenge training image in the quality inspection discrimination information sample set to obtain the microplate quality inspection discrimination network comprises: acquiring two quality inspection discrimination information samples and corresponding matching types from the quality inspection discrimination information sample set; According to the set strengthening weight, carrying out feature strengthening on the obtained knowledge features corresponding to the two quality inspection judging information samples and carrying out feature strengthening on the knowledge features corresponding to the corresponding two matching types, wherein the obtained result is used as a debugging training example; debugging the basic quality inspection discrimination network by using the obtained debugging training examples to obtain the micro-pore plate quality inspection discrimination network; The step of debugging the basic quality inspection discrimination network through the quality inspection discrimination information sample and the countermeasure training image in the quality inspection discrimination information sample set to obtain the micro-pore plate quality inspection discrimination network comprises the following steps: disassembling quality inspection discrimination information samples in the quality inspection discrimination information sample set into a plurality of training sample sets; For each training sample group, loading the training sample groups into a first initial decision tree network and a second initial decision tree network for debugging to obtain first debugging output information and second debugging output information, wherein the first initial decision tree network and the second initial decision tree network are obtained by initializing a basic quality inspection judging network; Acquiring quality inspection judging information samples with inconsistent decision information in the first debugging output information and the second debugging output information as quality inspection judging information samples to be optimized; Selecting a quality inspection judging information sample with a set proportion from the quality inspection judging information samples to be optimized as a first training set based on the training error of the quality inspection judging information samples to be optimized in the first debugging output information, and selecting the quality inspection judging information sample with the set proportion from the quality inspection judging information samples to be optimized as a second training set based on the training error of the quality inspection judging information samples to be optimized in the second debugging output information; Performing network parameter adjustment on the second initial decision tree network based on the first training set, and performing network parameter adjustment on the first initial decision tree network according to the second training set; And determining the first initial decision tree network and the second initial decision tree network as the microplate quality inspection discrimination network.
9. The method of claim 7, wherein the debugging the basic quality inspection discrimination network through the quality inspection discrimination information samples and the challenge training image in the quality inspection discrimination information sample set to obtain the microplate quality inspection discrimination network comprises: deriving quality inspection discrimination information samples and countermeasure training images in the quality inspection discrimination information sample set according to a plurality of set quality inspection discrimination matters to obtain a plurality of network training derivative samples corresponding to each quality inspection discrimination matter; Screening the corresponding network training derivative samples for each quality inspection discrimination item to obtain a network training derivative sample set; Executing processing of corresponding quality inspection discrimination items in a basic quality inspection discrimination network based on the network training derivative sample set to obtain debugging output information of each quality inspection discrimination item; And adjusting network parameters of the basic quality inspection discrimination network according to the debugging output information of each quality inspection discrimination item to obtain the micro-pore plate quality inspection discrimination network.
10. An image data processing system comprising at least one processor and a memory, the memory storing computer-executable instructions, the at least one processor executing the computer-executable instructions stored in the memory, causing the at least one processor to perform the method of any one of claims 1-9.

Description

Machine learning-based high-throughput data processing method and system for focused micro-pore plate Technical Field The invention relates to the technical field of data processing, in particular to a machine learning-based high-throughput data processing method and system for a focusing micro-pore plate. Background In the fields of biology, medicine and drug development, microplates are widely used as a common experimental tool, which can accommodate a plurality of independent microwells, and in each microwell can culture cells, store reagents or perform various biochemical reactions. As technology advances, high-throughput imaging techniques have been introduced into microplate processing to rapidly image a large number of microwells simultaneously, thereby producing high-throughput image data containing a large amount of experimental data. However, quality assessment of high throughput image data, especially during automated processing, still faces significant challenges. Conventional microplate quality inspection relies on manual visual inspection or simple image processing algorithms, which are not only inefficient, but also prone to omission or misinterpretation when processing high throughput image data. Manual inspection is affected by factors such as fatigue degree, experience and subjective judgment of people, stability and consistency of quality detection are difficult to ensure, and a simple image processing algorithm often lacks effective identification and discrimination capability when facing complex image background and changeable micro-pore plate types. Disclosure of Invention In order to improve the problems, the invention provides a machine learning-based high-throughput data processing method and system for a focusing micro-pore plate. In a first aspect, an embodiment of the present invention provides a machine learning-based high throughput data processing method for a focused micro-plate, which is applied to an image data processing system, and the method includes: acquiring a to-be-processed microplate image and a plurality of microplate quality detection viewpoints, wherein the to-be-processed microplate image comprises microplate type to be detected and high-throughput image data; Combining the quality detection viewpoints of the plurality of micro-pore plates and the high-throughput image data of the micro-pore plate image to be processed to generate initial quality inspection discrimination information; Performing microplate quality detection viewpoint decision on the initial quality detection discrimination information through a microplate quality detection discrimination network, and determining a microplate quality detection discrimination result of the to-be-processed microplate image, wherein the microplate quality detection discrimination result is one of the plurality of microplate quality detection viewpoints; The debugging step of the micro-pore plate quality inspection discrimination network comprises the following steps: obtaining X network training samples, wherein X is an integer greater than 0; For each network training sample, acquiring the type of the to-be-detected micro-plate of the current network training sample as a target micro-plate type and acquiring the type of the to-be-detected micro-plate of Y rest network training samples from X-1 network training samples as an anti-micro-plate type, wherein Y is an integer greater than 0 and less than the number of set viewpoints, and the number of set viewpoints is the maximum number of priori quality detection viewpoints in the micro-plate quality inspection discrimination results of micro-plate quality detection viewpoint decision; performing image combination according to the target micro-pore plate type, the anti-micro-pore plate type and the high-throughput image data of each network training sample to generate a quality inspection distinguishing information sample, and obtaining a quality inspection distinguishing information sample set, wherein the target micro-pore plate type is a matching type corresponding to the quality inspection distinguishing information sample, the quality inspection distinguishing information sample set comprises a plurality of quality inspection distinguishing information samples, and each quality inspection distinguishing information sample is correspondingly generated through one network training sample; And debugging the basic quality inspection discrimination network based on the quality inspection discrimination information sample set to obtain the micro-pore plate quality inspection discrimination network. Preferably, the generating initial quality inspection discrimination information based on the quality detection viewpoints of the plurality of microplates and the high-throughput image data of the to-be-processed microplates includes: Image combination is carried out on the quality detection viewpoints of the plurality of micro-pore plates and the high-flux image data of the