Waveform analyzer
A model constructed by a training process using the technique of deep learning using the training data including images created from a large number of chromatograms and correct peak information is previously stored in a trained model storage section. When chromatogram data for a target sample acquired with an LC measurement unit are inputted, an image creator converts the chromatogram into an image and creates an input image in which one of the two areas divided by the chromatogram curve as the boundary in the image is filled. A peak position estimator inputs the pixel values of the input image into a trained model using a neural network, and obtains the position information of the starting point and/or ending point of the peak and a peak detection confidence as the output. A peak determiner determines the starting point and/or ending point of each peak based on the peak detection confidence.
1. A waveform analyzer configured to analyze a signal waveform based on a signal sequence acquired by performing a predetermined analysis on a sample, so as to detect a peak on the signal waveform, the waveform analyzer comprising:
a) an image creator configured to create an input image by converting an original signal waveform or a secondary signal waveform derived from the original signal waveform into an image, dividing the image into first and second areas by a line corresponding to the signal waveform in the image, and filling at least a part of the first area, using one or more colors distinguishable from the second area; and
b) a peak detector configured to detect at least a position of a starting point or position of an ending point of one or more peaks appearing on a signal waveform in a target input image created by the image creator, using a trained model previously constructed by machine learning using input images created based on a plurality of reference signal waveforms whose starting points and ending points are known.
2. The waveform analyzer according to claim 1 , wherein:
the image creator is configured to generate one or more secondary signal waveforms from a single signal waveform acquired by an analysis, and create a plurality of input images based on the plurality of signal waveforms, respectively, or create an input image by superposing the plurality of input images on each other.
3. The waveform analyzer according to claim 2 , wherein:
the image creator is configured to create the input image by converting a signal waveform obtained by n-th order differentiation of an original signal waveform acquired by an analysis (where n is a positive integer) into an image along with the original signal waveform.
4. The waveform analyzer according to claim 1 , wherein:
the image creator is configured to create the input image by creating an image in which a plurality of signal waveforms acquired by performing a same analysis on a same sample under different values of a parameter are superposed on each other, then dividing the created image into three or more areas by a plurality of lines respectively corresponding to the signal waveforms, and filling each of the three or more areas with a different color.
5. The waveform analyzer according to claim 1 , wherein:
the image creator is configured to create the input image in which a plurality of signal waveforms acquired by repeating an analysis on a target sample a plurality of times are superposed on or combined with each other.
6. The waveform analyzer according to claim 1 , wherein:
the image creator is configured to create the input image in which a signal waveform acquired for a target sample and a signal waveform acquired for a standard sample containing a target component in the target sample are superposed on or combined with each other.
7. The waveform analyzer according to claim 1 , wherein:
the trained model is a model constructed by a general object detection algorithm using machine learning.
8. The waveform analyzer according to claim 7 , wherein:
the trained model is a model constructed by using deep learning.
9. The waveform analyzer according to claim 8 , wherein:
the trained model is a model constructed by using a convolutional neural network.
10. The waveform analyzer according to claim 2 , wherein:
the trained model is a model constructed by a general object detection algorithm using machine learning.
11. The waveform analyzer according to claim 3 , wherein:
the trained model is a model constructed by a general object detection algorithm using machine learning.
12. The waveform analyzer according to claim 4 , wherein:
the trained model is a model constructed by a general object detection algorithm using machine learning.
13. The waveform analyzer according to claim 5 , wherein:
the trained model is a model constructed by a general object detection algorithm using machine learning.
14. The waveform analyzer according to claim 6 , wherein:
the trained model is a model constructed by a general object detection algorithm using machine learning.
15. The waveform analyzer according to claim 1 , wherein:
the image creator is configured to create an input image by wholly filling the first area using one or more colors.