# US 10,368,804 B2 · Device, system and method for detection of fluid accumulation

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

- Number: US 10,368,804 B2
- Document ID: US10368804B2
- Status: Granted
- Kind: B2
- Application number: 14/421,505
- Filed: 2013-08-14
- Published: 2015-07-09
- Granted: 2019-08-06
- Earliest claimed: 2012-08-14
- Assignees: Nanyang Technological University
- Applicants: Nanyang Technological University
- Inventors: Wee Ser; Feng Yang; Junfeng Yu
- Web page: https://patentgrep.com/patent/US10368804B2

## CPC

- A61B 5/4875: {Hydration status, fluid retention of the body}
- G06F 18/00: Pattern recognition
- A61B 5/7225: {Details of analogue processing, e.g. isolation amplifier, gain or sensitivity adjustment, filtering, baseline or drift compensation (input circuits for detecting, measuring, or recording bioelectric or biomagnetic signals A61B5/30; specific diagnostic methods using bioelectric or biomagnetic signals A61B5/316)}
- A61B 5/726: {using Wavelet transforms}
- A61B 5/7267: {involving training the classification device}
- A61B 5/7275: {Determining trends in physiological measurement data; Predicting development of a medical condition based on physiological measurements, e.g. determining a risk factor}
- A61B 5/7278: {Artificial waveform generation or derivation, e.g. synthesizing signals from measured signals}
- A61B 7/04: Electric stethoscopes
- A61B 2560/0475: Special features of memory means, e.g. removable memory cards
- A61B 2562/0204: Acoustic sensors
- G06F 2218/00: Aspects of pattern recognition specially adapted for signal processing
- G06V 2201/03: Recognition of patterns in medical or anatomical images
- A61B: DIAGNOSIS; SURGERY; IDENTIFICATION (usefulness limited to only animals A61D)
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)
- G06V: IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING

## Abstract

A method of detecting a level of fluid accumulation in an internal organ of a subject is proposed, as well as a system for carrying out the method. The method comprises: providing at least one classifier trained to distinguish between two or more levels of fluid accumulation; acquiring an audio signal (110) generated by said internal organ; and processing, using at least one processor (134), said audio signal (110) by: performing feature extraction to generate at least one feature vector from the audio signal; and assigning a fluid level from the two or more levels to the audio signal by passing the at least one feature vector to the at least one classifier.

## Claims

### Claim 1

A method of detecting a level of fluid accumulation in a lung of a subject, comprising:
providing at least one classifier trained to distinguish between two or more levels of fluid accumulation, the two or more levels comprising a level corresponding to a presence of fluid and a level corresponding to an absence of fluid;
acquiring an audio signal generated by the lung;
processing said audio signal by:
segmenting the audio signal into consecutive time windows;
performing feature extraction to generate at least one feature vector for each time window; and
assigning a level of fluid accumulation from the two or more levels to the audio signal by:
passing at least one component of each feature vector to the at least one classifier to generate a plurality of tentative fluid levels for the audio signal; and
selecting a most frequent occurring level from amongst the plurality of tentative fluid levels; and
providing an indication to a user that the subject has lung fluid accumulation if the most frequent occurring level is the level corresponding to the presence of fluid.

### Claim 2 (depends on 1)

A method according to claim 1, wherein said consecutive time windows at least partially overlap.

### Claim 3 (depends on 1)

A method according to claim 1, wherein the passing at least one component of each feature vector to the at least one classifier comprises passing selected components of each feature vector to the at least one classifier.

### Claim 4 (depends on 3)

A method according to claim 3, wherein performing the feature extraction comprises determining mel-frequency cepstral coefficients or determining perceptual linear prediction coefficients, or both.

### Claim 5 (depends on 4)

A method according to claim 4, wherein determining the mel-frequency cepstral coefficients comprises selecting between 4 to 8 components.

### Claim 6 (depends on 4)

A method according to claim 4, wherein determining the perceptual linear prediction coefficients comprises selecting between 7 to 11 components.

### Claim 7 (depends on 1)

A method according to claim 1, wherein the at least one classifier comprises at least one linear classifier.

### Claim 8 (depends on 1)

A method according to claim 1, wherein the at least one classifier comprises at least one non-linear classifier.

### Claim 9 (depends on 1)

A method according to claim 1, further comprising training the at least one classifier.

### Claim 10 (depends on 9)

A method according to claim 9, wherein said training comprises:
acquiring audio signals from a plurality of subjects having known levels of fluid accumulation;
performing feature extraction on the audio signals acquired from the plurality of subjects having known levels of fluid accumulation to generate a plurality of candidate feature vectors; and
performing feature selection on the plurality of candidate feature vectors to identify one or more components of the candidate feature vectors which are useful for discriminating between the two or more levels.

### Claim 11 (depends on 10)

A method according to claim 10, wherein the feature selection is performed by identifying N features having greatest importance, from the plurality of candidate feature vectors and wherein an importance of a feature is calculated according to:
I
⁡
(
x
i
)
=
∑
j
≠
k
⁢
(
m
i
,
c
⁡
(
j
)
-
m
i
,
c
⁡
(
k
)
)
2
∑
c
⁡
(
j
)
⁢
σ
i
,
c
⁡
(
j
)
2
,
where i=1, . . . , M; M is a dimension of each candidate feature vector; mi,c(j) and σi,c(j)2 are a sample mean and a variance, respectively, of a feature xi for a class c(j);
c(j), j=1, 2, . . . Nc is a jth class; and Nc is a number of levels of fluid accumulation.

### Claim 12 (depends on 11)

A method according to claim 11, wherein M ranges from 10 to 15.

### Claim 13 (depends on 1)

A method according to claim 1, wherein said acquiring comprises positioning at least one sensor on the subject in proximity to the lung such that the sensor is acoustically coupled to the lung.

### Claim 14 (depends on 1)

A method according to claim 1, wherein the indication comprises an alarm signal.

### Claim 15 (depends on 1)

A method according to claim 1, wherein said acquiring comprises acquiring the audio signals using a plurality of audio sensors.

### Claim 16 (depends on 15)

A method according to claim 15, wherein said acquiring comprises positioning the plurality of audio sensors on the subject in proximity to the lung such that each of the sensors is coupled to the lung.

### Claim 17

A system for detecting a level of fluid accumulation in a lung of a subject, comprising:
at least one sensor positionable on the subject adjacent the lung to measure an audio signal generated by the lung;
a memory configured to store data representing the audio signal; and
at least one processor in communication with the memory and being configured to:
retrieve, from said memory, data representing at least one classifier trained to distinguish between two or more levels of fluid accumulation, the two or more levels comprising a level corresponding to a presence of fluid and a level corresponding to an absence of fluid;
process said audio signal by segmenting the audio signal into consecutive time windows, and performing feature extraction to generate at least one feature vector for each time window;
assign a level of fluid accumulation from the two or more levels to the audio signal by:
passing at least one component of each feature vector to the at least one classifier to generate a plurality of tentative fluid levels for the audio signal; and
selecting a most frequent occurring level from amongst the plurality of tentative fluid levels; and
provide an indication to a user that the subject has lung fluid accumulation if the most frequent occurring level is the level corresponding to the presence of fluid.

### Claim 18 (depends on 17)

A system according to claim 17 wherein the indication comprises an alarm signal.

### Claim 19

A system for detecting a level of fluid accumulation in a lung of a subject, comprising:
at least one sensor positionable on the subject adjacent the lung to measure an audio signal generated by the lung; and
at least one processor configured to:
segment the audio signal into consecutive time windows;
perform feature extraction to generate at least one feature vector for each time window;
assign to each time window a fluid level which is one of a set of two or more levels of fluid accumulation by passing at least one component of the at least one feature vector for each time window to at least one classifier trained to distinguish between the two or more levels of fluid accumulation to generate a plurality of tentative fluid levels for the audio signal, wherein the two or more levels comprises a level corresponding to a presence of fluid and a level corresponding to an absence of fluid;
select a most frequent occurring level from amongst the plurality of tentative fluid levels and
provide an indication to a user that the subject has lung fluid accumulation if the most frequent occurring level is the level corresponding to the presence of fluid.

### Claim 20 (depends on 19)

A system according to claim 19 wherein the indication comprises an alarm signal.