Evaluating input data using a deep learning algorithm

Patent metadata

Field	Value
Publication number	US-11842268-B2
Application number	US-201816648719-A
Country	US
Kind code	B2
Filing date	Sep 10, 2018
Priority date	Sep 20, 2017
Publication date	Dec 12, 2023
Grant date	Dec 12, 2023

Abstract

The invention provides a method for evaluating a set of input data, the input data comprising at least one of: clinical data of a subject; genomic data of a subject; clinical data of a plurality of subjects; and genomic data of a plurality of subjects, using a deep learning algorithm. The method includes obtaining a set of input data, wherein the set of input data comprises raw data arranged into a plurality of data clusters and tuning the deep learning algorithm based on the plurality of data clusters. The deep learning algorithm comprises: an input layer; an output layer; and a plurality of hidden layers. The method further includes performing statistical clustering on the raw data using the deep learning algorithm, thereby generating statistical clusters and obtaining a marker from each statistical cluster. Finally, the set of input data is evaluated based on the markers to derive data of medical relevance in respect of the subject or subjects.

First claim

The invention claimed is:

1. A computer-implemented method for evaluating a set of input data using a deep learning neural network, the input data comprising at least one of: clinical data of a first subject; genomic data of a second subject; clinical data of a first plurality of subjects; or genomic data of a second plurality of subjects, the method comprising:
obtaining a set of input data, wherein the set of input data comprises raw data arranged into a plurality of data clusters;
tuning the deep learning neural network based on the plurality of data clusters, wherein the deep learning neural network comprises:
an input layer;
an output layer; and
a plurality of hidden layers;

performing statistical clustering on the raw data using the deep learning neural network to generate statistical clusters;
determining that two raw data points of the raw data are clustered into different data clusters of the plurality of data clusters of the input data;
determining that the two raw data points are clustered into a same statistical cluster of the statistical clusters;
categorizing the two raw data points as an instance pair based on the two raw data points being clustered into the different data clusters of the plurality of data clusters of the input data, and being clustered in the same statistical cluster;
identifying vectors associated with the two raw data points of the instance pair;
calculating an angle between the vectors;
when the angle is above a threshold,
presenting the instance pair to a user;
obtaining a user input from the user, indicating whether the instance pair should be clustered together; and
tuning the deep learning neural network based on the user input;

obtaining a marker from each statistical cluster, wherein each marker is a biomarker relating to a single clinical parameter contained within a cluster of the statistical clusters; and
evaluating the set of input data based on the markers to derive data of medical relevance to the first subject, the second subject, the first plurality of subjects or the second plurality of subjects, wherein the markers are evaluated with reference to historical subject data collected from subjects with similar conditions and/or symptoms to determine a survival rate and/or an effective treatment method for the first subject, the second subject, the first plurality of subjects or the second plurality of subjects.

2. A method as claimed in claim 1 , wherein the tuning of the deep learning neural network comprises:
determining a Gaussian mean width of the input data;
determining a convergence rate of a loss-function of the deep learning neural network; and
selecting a number of hidden layers based on the Gaussian mean width and the convergence rate.

3. A method as claimed in claim 2 , wherein the determining of the Gaussian mean width is based on at least one of the size of the plurality of data clusters of the input data and a number of the plurality of data clusters of the input data.

4. A method as claimed in claim 1 , wherein the deep learning neural network is an autoencoder.

5. A method as claimed in claim 1 , wherein the performing of the statistical clustering comprises considering the hidden layers of the deep learning neural network as low dimensional representations, wherein principal variables of the hidden layers are identified to perform the statistical clustering in order to reduce a number of computations required.

6. A method as claimed in claim 1 , wherein the statistical clustering is k-means clustering.

7. A method as claimed in claim 1 , wherein the evaluating of the set of input data comprises predicting a survival rate of the first or second subject.

8. A method as claimed in claim 1 , wherein the statistical clustering is performed between adjacent hidden layers of the deep learning neural network.

9. A method as claimed in claim 1 , wherein the threshold to the angle is used to limit a number of instance pairs presented to the user.

10. A computer program comprising computer program code means which is adapted, when said computer program is run on a computer, to implement the method of claim 1 .

11. A controller for controlling an evaluation of a set of input data using a deep learning neural network, wherein the controller is adapted to:
obtain a set of input data, wherein the set of input data comprises raw data arranged into a plurality of data clusters, the input data comprising at least one of: clinical data of a first subject; genomic data of a second subject; clinical data of a first plurality of subjects; or genomic data of a second plurality of subjects;
tune the deep learning neural network based on the plurality of data clusters, wherein the deep learning neural network comprises:
an input layer;
an output layer; and
a plurality of hidden layers;

perform statistical clustering on the raw data using the deep learning neural network to generate statistical clusters;
determine that two raw data points of the raw data are clustered into different data clusters of the plurality of data clusters of the input data;
determine that the two raw data points are clustered into a same statistical cluster of the statistical clusters;
categorize the two raw data points as an instance pair based on the two raw data points being clustered into the different data clusters of the plurality of data clusters of the input data, and being clustered in the same statistical cluster;
identify vectors associated with the two raw data points of the instance pair;
calculate an angle between the vectors;
when the angle is above a threshold,
present the instance pair to a user;
obtain a user input from the user, indicating whether the instance pair should be clustered together; and
tune the deep learning neural network based on the user input;

obtain a marker from each statistical cluster, wherein each marker is a biomarker relating to a single clinical parameter contained within a cluster of the statistical clusters; and
evaluate the set of input data based on the markers, wherein the markers are evaluated with reference to historical subject data collected from subjects with similar conditions and/or symptoms to determine a survival rate and/or an effective treatment method for the first subject, the second subject, the first plurality of subjects or the second plurality of subjects.

12. The controller as claimed in claim 11 , wherein the controller is further adapted to:
determine a Gaussian mean width of the input data;
compute a convergence rate of a loss-function of the deep learning neural network; and
select a number of hidden layers based on the Gaussian mean width and the convergence rate.

13. A data analysis system, the system comprising:
a storage device, adapted to store the input data;
the controller as claimed in claim 11 in communication with the storage device;
a user interface, adapted to obtain a user input, in communication with the controller; and
a display device, adapted to display information to a user, in communication with the controller.

Assignees

• Koninklijke Philips Nv

Inventors

• Mavroeidis Dimitrios
• Hendriks Monique
• Vos Pieter Christiaan
• Consoli Sergio
• Kustra Jacek Lukasz
• Janssen Johan
• Hoffmann Ralf Dieter

Classifications

• G06N3/0455

  Auto-encoder networks; Encoder-decoder networks · CPC title

• G06N3/091

  Active learning · CPC title

• G06N3/0499

  Feedforward networks · CPC title

• G06N3/08 Primary

  Learning methods · CPC title

• G06F17/18

  for evaluating statistical data {, e.g. average values, frequency distributions, probability functions, regression analysis (forecasting specially adapted for a specific administrative, business or logistic context G06Q10/04)} · CPC title