A method for automatic classification of immunofixation electropherograms based on their extracted features

Abstract

The invention discloses a method for automatically classifying the features of the extracted immunofixation electrophoresis images, which belongs to the technical field of machine learning and deep learning, and solves the problem of misjudgment easily caused by manually judging the IFE images in the prior art. The invention reads the IFE map to be classified and divides it, based on the input size of the start position prediction model and the end position prediction model of the target area, performs scaling processing on each segmented electrophoretic zone image, and obtains the processed Each electrophoretic zone image of each electrophoretic zone image; after the target region position of each electrophoretic zone image after the prediction process is predicted based on the start position prediction model and the end position prediction model of the target region, each target region image is extracted; based on the convolutional neural network model, the The image of the target area in each electrophoretic zone image is processed to obtain the final features, and the final features are spliced ​​to obtain the features of the IFE map to be classified, and then the classification result is obtained based on the IFE classification model. The present invention is used to classify IFE maps.

Description

technical field [0001] A method for automatically classifying immunofixed electrophoresis images based on extracted features is used for IFE image classification and belongs to the technical field of machine learning and deep learning. Background technique [0002] Immunofixation electrophoresis (IFE) refers to the separation of serum proteins by electrophoresis, and each sample reacts with a specific antibody for a specific heavy chain or light chain. Commonly used antibodies include anti-γ, d, μ and anti- For K and λ antibodies, the bands of immunoprecipitation reaction with the antibodies will be stained to obtain the IFE immunofixation electrophoresis. IgG, IgM, IgA, etc., as well as kappa light chain and lambda light chain can be detected by serum IFE. At present, it is generally accepted that IFE technology has the characteristics of high sensitivity and good specificity. [0003] Although IFE technology plays an irreplaceable advantage in classification, its automat...

AI Technical Summary

Problems solved by technology

[0005] In view of the problems of the above research, the purpose of the present invention is to provide a method for automatically classifying it based on the features of the extracted immunofixation electrophoresis, which solves the problem of using a deep learning model to directly extract the features of the IFE map in the prior art, and the representation of its features Insufficient ability; IFE technology is used to classify IFE diagrams, and its automation and standardization are poor; manual judgment of IFE diagrams is used to classify, which requires high personnel requirements, is time-consuming and labor-intensive, and has large deviations in classification results; Using the multi-classification model that can train IFE graphs in the deep learning model, the classification effect is not good

Images