# US 2019/0317118 A1 · Chemical sensing system

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 2019/0317118 A1
- Document ID: US20190317118A1
- Status: Published application
- Kind: A1
- Application number: 16/456,244
- Filed: 2019-06-28
- Published: 2019-10-17
- Earliest claimed: 2018-01-29
- Assignees: Stratuscent Inc.
- Applicants: Stratuscent Inc.
- Inventors: Mojtaba Khomami Abadi; Amir Bahador Gahroosi; Matthew V. Gould; Ashok Prabhu Masilamani
- Web page: https://patentgrep.com/patent/US20190317118A1

## CPC

- G01N 35/00722: {Communications; Identification}
- G06N 3/088: Non-supervised learning, e.g. competitive learning
- B82Y 15/00: Nanotechnology for interacting, sensing or actuating, e.g. quantum dots as markers in protein assays or molecular motors
- G01N 33/0031: {comprising two or more sensors, e.g. a sensor array}
- G01N 33/0034: {comprising neural networks or related mathematical techniques}
- G06N 3/0442: characterised by memory or gating, e.g. long short-term memory [LSTM] or gated recurrent units [GRU]
- G06N 3/045: Combinations of networks
- G06N 3/0455: Auto-encoder networks; Encoder-decoder networks
- G06N 3/0464: Convolutional networks [CNN, ConvNet]
- G06N 3/0475: Generative networks
- G06N 3/0895: Weakly supervised learning, e.g. semi-supervised or self-supervised learning
- G06N 3/09: Supervised learning
- G06N 3/094: Adversarial learning
- G06N 3/096: Transfer learning
- G01N 2035/00891: {Displaying information to the operator}
- G06N 3/044: Recurrent networks, e.g. Hopfield networks
- G01N: INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES (measuring or testing apparatus or processes other than immunoassay, involving enzymes or microorganisms C12M, C12Q)
- G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- B82Y: SPECIFIC USES OR APPLICATIONS OF NANOSTRUCTURES; MEASUREMENT OR ANALYSIS OF NANOSTRUCTURES; MANUFACTURE OR TREATMENT OF NANOSTRUCTURES

## Abstract

A chemical sensing system is described. The chemical sensing system can include a plurality of sensors arranged on at least one substrate. The sensors may have differing sensitivities to sense different analytes, and may each be configured to output a signal in response to sensing one or more of the different analytes. The chemical sensing system can further include a computer processor programmed to receive the signals output from the plurality of sensors. The computer processor may be further programmed to determine a concentration of analytes in the sample based, at least in part, on the received signals and a model relating the signals or information derived from the signals to an output representation having bases corresponding to analytes.

## Claims

### Claim at position 1

1-30. (canceled)

### Claim 31

A chemical sensing system, comprising:
a plurality of sensors arranged on at least one substrate, wherein a first sensor and a second sensor of the plurality of sensors have different sensitivities to sense at least one analyte in a sample, each of the plurality of sensors being configured to output a signal in response to sensing the at least one analyte; and
a computer processor programmed to:
receive the signals output from the plurality of sensors; and
determine a concentration of the at least one analyte in the sample based, at least in part, on the received signals and a model relating the signals or information derived from the signals to an output representation having bases corresponding to analytes, the analytes including the at least one analyte.

### Claim 32 (depends on 31)

The chemical sensing system of claim 31, wherein:
the model comprises:
a first model relating the signals output from the plurality of sensors to a feature representation; and
a second model relating the feature representation to the output representation.

### Claim 33 (depends on 32)

The chemical sensing system of claim 32, wherein:
the first model comprises:
a mapping from the signals to response values; and
a weighting relating the response values to the feature representation.

### Claim 34 (depends on 33)

The chemical sensing system of claim 33, wherein:
the weighting comprises:
weights of a recurrent neural network or convolutional neural network that relate the response values to the feature representation.

### Claim 35 (depends on 33)

The chemical sensing system of claim 33, wherein:
the mapping assigns to each response value an amplitude characteristic and/or a temporal characteristic of a corresponding signal.

### Claim 36 (depends on 35)

The chemical sensing system of claim 35, wherein:
the amplitude characteristic comprises a normalized change in an amplitude of the corresponding signal.

### Claim 37 (depends on 35)

The chemical sensing system of claim 35, wherein:
the temporal characteristic comprises a rise time or a fall time of an amplitude of the corresponding signal.

### Claim 38 (depends on 32)

The chemical sensing system of claim 32, wherein:
a change in a location of a point in the feature representation corresponding to an analyte concentration in the sample is a non-linear function of a change in the analyte concentration in the sample.

### Claim 39 (depends on 32)

The chemical sensing system of claim 32, wherein:
the second model relating the feature representation to the output representation comprises:
a first sub-model relating the feature representation to a latent representation, the latent representation having a lower dimensionality than the feature representation.

### Claim 40 (depends on 32)

The chemical sensing system of claim 32, wherein:
the second model relating the feature representation to the output representation comprises:
a second sub-model relating a latent representation to a straightened orthogonal representation, wherein:
one-dimensional manifolds in the straightened orthogonal representation corresponding to varying concentrations of a same analyte have zero angle between them; and
one-dimensional manifolds in the straightened orthogonal representation corresponding to different analytes are orthogonal.

### Claim 41 (depends on 32)

The chemical sensing system of claim 32, wherein:
the second model relating the feature representation to the output representation comprises:
a third sub-model relating a straightened orthogonal representation to the output representation, wherein the third sub-model:
relates the straightened orthogonal representation to a non-linearized output representation having standard bases associated with the analytes; and
relates the non-linearized output representation to the output representation; and
wherein a change in a location of a point in the output representation corresponding to an analyte concentration in the sample is a linear function of a change in the analyte concentration in the sample.

### Claim 42 (depends on 32)

The chemical sensing system of claim 32, wherein:
each signal comprises a time series of measurements.

### Claim 43 (depends on 31)

The chemical sensing system of claim 31, wherein:
the model comprises:
a device-specific model relating the signals or information derived from the signals to an intermediate representation; and
a device-independent model relating the intermediate representation to the output representation.

### Claim 44

A chemical sensing system, comprising:
at least one computer processor; and
at least one computer readable medium including instructions that, when executed by the at least one computer processor, cause the chemical sensing system to perform a training process, the training process comprising:
receiving a training dataset generated by a plurality of sensors having different sensitivities to sense at least one analyte in a sample, each of the plurality of sensors configured to output a signal in response to sensing the at least one analyte;
training, using the training dataset, an assessment model relating the signals and/or information derived from the signals to an output representation having bases corresponding to analytes, the analytes including the at least one analyte in the sample; and
configuring the chemical sensing system to detect concentrations of the analytes using the trained assessment model relating the signals or information derived from the signals to the output representation.

### Claim 45 (depends on 44)

The system of claim 44, wherein:
training the assessment model comprises:
generating a first model relating the signals output from the plurality of sensors to a feature representation; and
generating a second model relating the feature representation to the output representation.

### Claim 46 (depends on 45)

The system of claim 45, wherein:
generating the first model comprises training weights in a recurrent neural network or convolutional neural network using the training dataset.

### Claim 47 (depends on 45)

The system of claim 45, wherein:
generating the first model comprises training the model using an unsupervised learning technique.

### Claim 48 (depends on 45)

The system of claim 45, wherein:
generating the second model relating the feature representation to the output representation comprises generating a first sub-model relating the feature representation to a latent representation having a dimensionality less than the dimensionality of the feature representation.

### Claim 49 (depends on 48)

The system of claim 48, wherein:
generating the first sub-model comprises training weights in an autoencoder or generative adversarial network using the training dataset.

### Claim 50 (depends on 48)

The system of claim 48, wherein:
generating the first sub-model comprises training the first sub-model using an unsupervised learning technique.

### Claim 51 (depends on 48)

The system of claim 48, wherein:
a continuous manifold embedded in the feature representation contains the signals output from the plurality of sensors, the continuous manifold having a dimensionality equal to the dimensionality of the latent representation; and
generating the first sub-model comprises relating the continuous manifold to the latent representation.

### Claim 52 (depends on 48)

The system of claim 48, wherein:
training the assessment model further comprises automatically identifying manifolds corresponding to samples of varying concentrations of a same analyte using local gradient changes in the feature representation.

### Claim 53 (depends on 48)

The system of claim 48, wherein:
generating the second model relating the feature representation to the output representation further comprises generating a second sub-model relating the latent representation to a straightened, orthogonal representation.

### Claim 54 (depends on 53)

The system of claim 53, wherein:
generating the second sub-model comprises:
training at least one neural network using the training dataset and at least one loss function, the at least one loss function configured to:
penalize non-zero angles between vectors output by the at least one neural network that correspond to samples of varying concentrations of a same analyte; and
penalize non-orthogonal vectors output by the at least one neural network that correspond to samples of different analytes.

### Claim 55 (depends on 54)

The system of claim 54, wherein:
the at least one neural network comprises:
a first feedforward neural network that relates the latent representation to a straightened representation; and
a second feedforward neural network that relates the straightened representation to a straightened, orthogonal representation.

### Claim 56 (depends on 55)

The system of claim 55, wherein:
training the at least one feedforward neural network comprises training the first feedforward neural network and/or the second feedforward neural network using an unsupervised learning technique.

### Claim 57 (depends on 53)

The system of claim 53, wherein:
generating the second model relating the feature representation to the output representation further comprises associating bases in the straightened orthogonal representation with the analytes.

### Claim 58 (depends on 53)

The system of claim 53, wherein:
generating the second model relating the feature representation to the output representation further comprises generating a third sub-model that relates the straightened orthogonal representation to the output representation.

### Claim 59 (depends on 53)

The system of claim 53, wherein:
generating the second model relating the feature representation to the output representation further comprises:
associating bases in the straightened orthogonal representation with the analytes using labels in the training data;
generating an introduction model that introduces the associated bases onto the standard bases of a non-linearized output representation; and
generating a linearizing model that linearizes the non-linearized output representation using concentrations in the training data.

### Claim 60 (depends on 59)

The system of claim 59, wherein:
the linearizing model that linearizes the non-linearized output representation comprises at least one feedforward neural network trained to linearize the non-linearized output representation using the training data and concentration values corresponding to the training data.

### Claim 61

A chemical sensing system, comprising:
a plurality of sensors arranged on at least one substrate, wherein a first sensor and a second sensor of the plurality of sensors have different sensitivities to sense at least one analyte in a sample, each of the plurality of sensors being configured to output a signal in response to sensing the at least one analyte; and
a computer processor programmed to:
receive the signals output from the plurality of sensors; and
determine a location in a latent representation using a homeomorphic map relating the signals or information derived from the signals to locations in the latent representation;
identify locations in the latent representation in a neighborhood of the determined location; and
determine concentrations of one or more analytes in the sample using known concentrations of one or more analytes corresponding to the identified data points.

### Claim 62 (depends on 61)

The chemical sensing system of claim 61, wherein:
the concentrations of the one or more analytes in the sample are determined by interpolation of the known concentrations.

### Claim 63

A chemical sensing system, comprising:
a plurality of sensors arranged on at least one substrate, wherein a first sensor and a second sensor of the plurality of sensors have different sensitivities to sense at least one analyte in a sample, each of the plurality of sensors being configured to output a signal in response to sensing the at least one analyte; and
a computer processor programmed to:
receive the signals output from the plurality of sensors;
when the received signals and stored signals satisfy an update criterion;
update a first model relating the signals output from the plurality of sensors to a feature representation;
update a second model relating the feature representation to a latent representation; and
configure the chemical sensing system to determine concentrations of one or more analytes using the updated first model and the updated second model.

### Claim 64 (depends on 63)

The chemical sensing system of claim 63, wherein:
the update criterion depends on a value of mutual information between the signals and the stored signals.

### Claim 65

A method comprising:
receiving signals from a plurality of sensors of a chemical sensing system, the sensors having different sensitivities to sense at least one analyte in the sample, each of the sensors configured to output a signal in response to sensing the at least one analyte;
mapping the responses to an output representation having coordinate bases corresponding to analytes using a homeomorphic mapping; and
determining a concentration of a chemical in the sample from the output representation.

### Claim 66

A training method comprising:
receiving a training dataset generated by a plurality of sensors of a chemical sensing system, the sensors having different sensitivities to sense at least one analyte in a sample, each of the sensors configured to output a signal in response to sensing the at least one analyte;
generating a first model relating the signals output from the plurality of sensors to a feature representation using the training dataset; and
generating a second model relating the feature representation to an output representation having coordinate bases corresponding to analytes using the training dataset; and
configuring a chemical sensing system to detect concentrations of the analytes using the first model and the second model.
