JPWO2020097302A5

JPWO2020097302A5 -

Info

Publication number: JPWO2020097302A5
Application number: JP2021525318A
Authority: JP
Publication date: 2022-09-22
Anticipated expiration: 2039-11-07

Claims

obtaining a set of images associated with a surface using an atomic force microscope; combining the images; using a machine learning method applied to the combined images; classifying the surface, the method comprising:
using the atomic force microscope includes using a multi-channel atomic force microscope;
each channel corresponding to a surface attribute of said surface;
classifying the surface includes classifying the use of a machine learning module that is trained using training data, testable using test data, and verifiable using validation data;
The training data is data that has been used to learn a classification method, and the test data is data that can be used to validate the classification performed by the machine learning module. .

2. The method of claim 1, further comprising processing the image to obtain surface parameters and using machine learning to classify the surface based at least in part on the surface parameters. .

2. The method of claim 1, further comprising selecting the surface as being the surface of cells collected from urine of a subject, and classifying the cells as exhibiting cancer or not exhibiting cancer. the method of.

2. The method of claim 1, further comprising condensing information provided by said channels and building a condensed database from said condensed information.

further comprising condensing information provided by said channels and building an agglomerated database from said condensed information;
Building the condensed database includes:
forming a first database based on the information provided by the channel, wherein the first database has an index;
deriving a condensed database from the first database, wherein the condensed database has fewer indexes than the first database;
including
Deriving the condensed database includes performing tensor addition to combine information from the first database with one or more slices corresponding to one or more indices of the first database. 2. The method of claim 1, comprising generating sums and using the tensor sums to form the condensed database.

further comprising condensing information provided by said channels and building an agglomerated database from said condensed information;
Building the condensed database includes:
forming a first database based on the information provided by the channel, wherein the first database has an index;
deriving a condensed database from the first database, wherein the condensed database has fewer indexes than the first database;
including
Deriving a condensed database from the first database comprises:
defining a subset of values from said first database, wherein each of said values represents a corresponding element in said first database;
deriving a condensed value from the values in the subset of values;
representing the corresponding element from the first database using the condensed value
and
2. The method of claim 1, wherein deriving the condensation value comprises passing information from the first database through a surface parameter extractor to obtain a surface parameter set.

the image is an image of a sample;
2. The method of claim 1, further comprising dividing an image of the sample into sections, obtaining a surface parameter for each section, and defining the surface parameter of the image as the median or average of the surface parameters of each section. The method described in .

2. The method of claim 1, further comprising processing the image to obtain surface parameters, and using machine learning to classify the surface from externally derived parameters based at least in part on the surface parameters. The method described in .

said surfaces are surfaces of bodies derived from collected samples, at least one of said samples being a body-free sample without bodies;
processing the image to obtain surface parameters and using machine learning to classify the surface from externally derived parameters based at least in part on the surface parameters;
selecting the externally derived parameter as including data indicative of the absence of body from the body-free sample;
2. The method of claim 1, further comprising:

the surface is the surface of a cell derived from a sample obtained from a patient;
processing the image to obtain surface parameters and using machine learning to classify the surface from externally derived parameters based at least in part on the surface parameters;
2. The method of claim 1, further comprising selecting the externally derived parameters as containing data indicative of the probability that the patient will have a particular disease.

classifying the surface based at least in part on the surface parameters from the acquired image using machine learning;
defining a subset of the surface parameters;
generating a database based on said subset;
further comprising
Defining the subset of surface parameters comprises: identifying correlations between the surface parameters; comparing the correlations to a threshold to identify a set of correlation parameters; and including a subset in said subset of surface parameters.

said surface is a surface of a first plurality of said cells from a patient, a second plurality of said cells classified as from a patient with cancer, and a third plurality of said cells from a non-cancer patient are classified as those of
2. The method of claim 1, further comprising diagnosing the patient as having cancer if the ratio of the second plurality to the first plurality exceeds a predetermined threshold.

The atomic force microscope includes a cantilever and a probe arranged at the tip of the cantilever, the cantilever has a resonance frequency, and using the atomic force microscope includes the probe and the 2. The method of claim 1, comprising vibrating the distance between the surfaces at a frequency below the resonant frequency.

2. The method of claim 1, further comprising automatically detecting from said images that images have artifacts, and automatically excluding said images having said artifacts from being used to classify said surfaces. described method.

2. The method of claim 1, further comprising selecting the surface to include one of a painting, currency, stock certificate, identity card, birth certificate, passport.

condensing information provided by the channel;
building a condensed database from the condensed information;
forming a first database based on the information provided by the channel;
further comprising
Constructing the condensed database includes projecting the first database into a lower dimensional subspace than the first database, the projection defining the condensed database, the condensed database comprising: 2. The method of claim 1, having a lower dimensionality than the first database.

condensing information provided by the channel;
building a condensed database from the condensed information;
forming a first database based on the information provided by the channel;
further comprising
Constructing the condensed database includes projecting the first database into a lower dimensional subspace than the first database, the projection defining the condensed database, the condensed database comprising: having a lower dimensionality than the first database;
2. The method of claim 1, wherein building the condensed database comprises including only surface parameters with the highest rankings based on their respective Gini indices.

processing the image to obtain surface parameters;
using machine learning to classify the surface from externally derived parameters based at least in part on the surface parameters, wherein the surface is the surface of a cell derived from a sample obtained from a patient; be,
selecting the externally derived parameter as including data indicative of the probability that the patient has a particular disease;
further comprising
2. The method of claim 1, wherein the data indicative of the probability includes the patient's age, the patient's smoking habits, and the patient's family history.

condensing information provided by the channel;
forming a first database based on the information provided by the channel;
projecting the first database into a lower dimensional subspace than the first database, wherein the projection defines the condensed database having a lower dimensionality than the first database;
further comprising
2. The method of claim 1, wherein classifying the surface comprises a method selected from the group consisting of random forest methods, extreme randomized forest methods, gradient boosting tree methods, using neural networks and decision tree methods. .

a multi-channel atomic force microscope for acquiring images associated with a surface , wherein each channel of the microscope corresponds to a surface attribute of the surface;
a processing system for receiving signals representing said images from said atomic force microscope and for combining said images , wherein said processing system comprises a machine learning module, and after learning a classification basis from said machine learning module , and a classifier for classifying unknown samples,
A device comprising
the machine learning module is trained to classify the surface using training data such that the classification results are verifiable using test data;
The apparatus, wherein the training data is data that has been used to learn a classification method and the test data is data that can be used to verify the effectiveness of the classification.