Cell division identification method

Abstract

The invention discloses a cell division identification method. The method comprises the following steps of: learning a first optimum dictionary through a first training set and a target function; sparsely decomposing the first training set through the first optimum dictionary, and acquiring first optimum sparse decomposition coefficients which correspond to all first vision characteristic vectors; acquiring a trained dividing cell model through the first optimum sparse decomposition coefficients which correspond to a positive sample and a negative sample; and acquiring second optimum sparse decomposition coefficients of new test data, inputting the second optimum sparse decomposition coefficients into the trained dividing cell model, and acquiring an output result, wherein the type of the output result is marked as 1, the new test data comprise a dividing cell area, and when the type of the output result is marked as 0, the new test data do not comprise the dividing cell area. By adoption of the cell division identification method, the difficulty in target characteristic extraction of a non-rigid body is overcome, identification of a cell division behavior is independent of cell tracking and time sequence deduction models, influence of computation complexity can be remarkably reduced, and the identification rate of cell division is improved.

Description

technical field [0001] The invention belongs to the field of image analysis and pattern recognition, in particular to a cell division recognition method. Background technique [0002] Quantitative analysis of cell proliferation behavior by means of in vitro culture is of great significance for many biomedical applications, such as drug development, stem cell culture, tissue engineering, etc. During the research process, cell image acquisition through a microscope and the use of advanced image processing techniques to achieve accurate cell division identification will play a key role in this research. [0003] Traditional cell division recognition methods based on microscope images can be divided into three categories: 1) Methods based on local saliency features: This method regards the cell division region as a special visual pattern, and uses the image features of the division region to directly learn a supporting The vector machine classifier is used for identification; 2...

AI Technical Summary

Problems solved by technology

[0005] 1) The existing visual features have limited ability to represent non-rigid and deformed objects (such as cells), so methods based on local salient features usually have poor generalization ability; 2) Tracking-based methods and time-series model-based methods are both There is a strong dependence on accurate trajectory extraction, however, the tracking of non-rigid objects is inherently difficult; 3) Time-series model-based methods often perform complex models through the utilization of a wide range of time-series information and the learning of time-series state transitions This will make the learning of the model require high computational complexity, and make this method unable to meet the real-time requirements of biological analysis for cell division recognition, and the recognition rate of cell division is low

Images