Fluorescence immunochromatography quantitative image peak searching algorithm based on deep learning

Abstract

The invention provides a fluorescence immunochromatography quantitative image peak searching algorithm based on deep learning. The fluorescence immunochromatography quantitative image peak searching algorithm comprises the following steps: collecting a large amount of fluorescence immunochromatography quantitative image data; labeling peak point positions in the collected fluorescence immunochromatography quantitative images to obtain label information of the image data; performing standardized preprocessing on the label information, and establishing an algorithm training set; establishing a first-layer convolutional neural network of a cascade algorithm, and positioning a peak point in a very small error range; a second-layer convolutional neural network of the cascade algorithm is established, so that the result is more accurate; and performing standardized preprocessing on the data of the test set, establishing the test set, inputting the test set into the trained algorithm network,and testing the peak searching accuracy of the fluorescence immunochromatography image. According to the fluorescence immunochromatography quantitative image peak searching algorithm based on deep learning, correct peak points can be recognized, and accurate peak point coordinate data can be output.

Description

technical field [0001] The invention belongs to the field of image processing, and in particular relates to a quantitative image peak-finding algorithm of fluorescence immunochromatography based on deep learning. Background technique [0002] The peak-finding methods currently used include direct peak-finding method, half-peak peak-finding method, general polynomial fitting method, Monte Carlo algorithm, Gaussian-polynomial fitting method and three-point peak-finding algorithm. The direct peak-finding method and the half-peak peak-finding method mainly use the first-order numerical differentiation method. The peak-finding method is realized by differentiating the global image. The calculation process is simple, but this method is only suitable for finding isolated peaks. For curves with large fluctuations The peak-finding accuracy of complex images is too low; the general polynomial fitting method, which uses general polynomials for fitting and uses the least square method f...

AI Technical Summary

Problems solved by technology

The direct peak-finding method and the half-peak peak-finding method mainly use the first-order numerical differentiation method. The peak-finding method is realized by differentiating the global image. The calculation process is simple, but this method is only suitable for finding isolated peaks. For curves with large fluctuations The peak-finding accuracy of complex images is too low; the general polynomial fitting method, which uses general polynomials for fitting and uses the least square method for judgment, has the advantage of being simple and easy to implement, but the peak-finding accuracy is low; Monte Carlo algorithm, Also known as the centroid detection method, it is a statistical simulation algorithm that has the advantage of fast calculation speed, but its algorithm linearity is not ideal, resulting in low peak finding accuracy; Gaussian-polynomial fitting method, Gaussian polynomial conversion of the fluorescence peak map For peak finding, the peak finding accuracy of this method is higher than that of the general polynomial fitting method, but this method has higher sensitivity to waveform images, weaker anti-noise ability, higher requirements on image peak shape, and is not affected by interference peaks. The accuracy of peak shape recognition is low; finally, regarding the three-point peak-finding algorithm, the peak-finding accuracy of this type of algorithm has been significantly improved, but there are strict requirements for the normalization of image peak shape

[0003] Generally speaking, the existing methods cannot accurately locate the peak points of images containing interference peaks. To solve the above problems, this paper proposes a convolutional neural network algorithm with cascading ideas, and extracts image features through a multi-level algorithm network. Use algorithms of different structures for feature mining and training on data at different stages, and obtain accurate peak finding results through parallel network prediction

Images