U.S. patent application number 14/210026 was filed with the patent office on 2014-09-18 for automated diagnosis-assisting medical devices utilizing pattern localization of quasi-periodic signals.
The applicant listed for this patent is Andreas J. Reinisch, Andreas J. Schriefl. Invention is credited to Andreas J. Reinisch, Andreas J. Schriefl.
Application Number | 20140275809 14/210026 |
Document ID | / |
Family ID | 50774833 |
Filed Date | 2014-09-18 |
United States Patent
Application |
20140275809 |
Kind Code |
A1 |
Schriefl; Andreas J. ; et
al. |
September 18, 2014 |
Automated Diagnosis-Assisting Medical Devices Utilizing Pattern
Localization Of Quasi-Periodic Signals
Abstract
A method for localizing a pattern in a quasi-periodic signal
includes estimating, using a controller, a rate or a frequency of a
quasi-periodic signal, and defining a search window based on the
estimated rate or frequency of the quasi-periodic signal. A
starting position is defined in the received quasi-periodic signal,
the starting position corresponding to a first maximum. A portion
of the quasi-periodic signal in the search window is
cross-correlated with a template signal pattern to be matched to
produce a second maximum. The second maximum is defined by the
controller as a new starting position. The new starting position is
stored.
Inventors: |
Schriefl; Andreas J.; (Graz,
AU) ; Reinisch; Andreas J.; (Graz, AU) |
|
Applicant: |
Name |
City |
State |
Country |
Type |
Schriefl; Andreas J.
Reinisch; Andreas J. |
Graz
Graz |
|
AU
AU |
|
|
Family ID: |
50774833 |
Appl. No.: |
14/210026 |
Filed: |
March 13, 2014 |
Related U.S. Patent Documents
|
|
|
|
|
|
Application
Number |
Filing Date |
Patent Number |
|
|
61787998 |
Mar 15, 2013 |
|
|
|
Current U.S.
Class: |
600/300 |
Current CPC
Class: |
A61B 5/024 20130101;
A61B 5/7282 20130101; A61B 5/0002 20130101; A61B 5/0205 20130101;
A61B 5/0022 20130101; A61B 5/742 20130101; A61B 5/7225 20130101;
A61B 7/003 20130101; A61B 5/7246 20130101; A61B 5/0816 20130101;
A61B 5/7405 20130101; A61B 7/04 20130101; A61B 5/7203 20130101;
G06K 9/0053 20130101; A61B 5/7235 20130101; A61B 5/72 20130101 |
Class at
Publication: |
600/300 |
International
Class: |
A61B 5/00 20060101
A61B005/00; A61B 7/04 20060101 A61B007/04 |
Claims
1. A method for localizing a pattern in a quasi-periodic signal,
the method comprising: estimating, using a controller, a rate or a
frequency of a quasi-periodic signal; defining, using the
controller, a search window based on the estimated rate or
frequency of the quasi-periodic signal; defining, using the
controller, a starting position in the received quasi-periodic
signal, the starting position corresponding to a first maximum;
cross-correlating, using the controller, a portion of the
quasi-periodic signal in the search window with a template signal
pattern to be matched to produce a second maximum that is defined
by the controller as a new starting position; and storing, using
the controller, the new starting position.
2. The method of claim 1, further comprising: determining, using
the controller, a localized pattern of the template signal pattern;
extracting, using the controller, a signal segment from the
localized pattern; analyzing, using the controller, the signal
segment simultaneously in time and frequency domains to produce
parallel outputs; combining, using the controller, the parallel
outputs via a statistical or mathematical function to produce a
result; and automatically indicating, using the controller, a
diagnosis based on the result.
3. The method of claim 2, further comprising indicating the
diagnosis directly on a device selected from a group consisting of
the electronic stethoscope, a second device attached to the
electronic stethoscope, and a third device in wireless
communication with the electronic stethoscope.
4. The method of claim 2, further comprising indicating the
diagnosis via one or more of a display and audio output of a
portable device.
5. The method of claim 1, further comprising: receiving, using the
controller, a representation of the quasi-periodic signal;
removing, using the controller, a DC component from the received
signal to produce a purely time-varying signal; filtering, using
the controller, the time-varying signal to produce a pre-processed
signal; auto-correlating, using the controller, at least a portion
of a representation of the pre-processed signal with itself and
storing in a memory device a corresponding auto-correlation output
for the at least portion of the representation of the pre-processed
signal; applying, using the controller, a biphasic tapering
function to the auto-correlation output, the tapering function
including a time constant parameter that is a function of the
quasi-periodic signal and producing the first maximum; and storing
in the memory device a representation, based on the first maximum,
as an indication of a rate or a frequency of the quasi-periodic
signal.
6. The method of claim 1, further comprising: saving, using the
controller, the representation in a file of a predetermined file
format; sending, using a data communication, the file in an e-mail;
printing, using a printer, the file; and storing the file on a
portable device.
7. The method of claim 1, further comprising: receiving, from an
auscultation device, the quasi-periodic signal as an auscultation
signal; recording, using the controller, raw auscultation data
associated with the auscultation signal; sending, using a data
connection, the raw data auscultation data to a hospital
information system; analyzing, using computation services of the
hospital information system, the raw auscultation data; and
storing, on a memory device of the hospital information system,
analysis results for later review by a medical professional.
8. The method of claim 1, further comprising: retrieving, using the
controller, a patient list from a hospital information system;
selecting, using a selection input, a patient from the patient
list; obtaining, using the controller, patient specific parameters;
and transferring to the hospital information system, via a data
connection, one or more of patient data, raw auscultation data, the
patient specific parameters, and a diagnosis suggestion.
9. The method of claim 1, further comprising: receiving, from an
auscultation device, the quasi-periodic signal as an auscultation
signal; recording, using the controller, auscultation data
associated with the auscultation signal; providing, using inputs of
the auscultation device, user setting options for modifying
settings of the auscultation device; communicating, using an audio
output of the auscultation device, between a medical professional
and an user of the auscultation device; alerting, using the
controller, the medical professional when the auscultation data is
uploaded and available for access; allowing, using the controller,
the medical professional to access the auscultation data.
10. The method of claim 1, further comprising: sending, via a data
connection, raw data of the rate or the frequency of the
quasi-periodic signal to a hospital information system; analyzing,
using another controller of the hospital information system, the
raw data; and storing, on a memory device of the hospital
information system, results of the analysis for later review by a
medical professional.
11. A system for localizing a pattern in a quasi-periodic signal,
the system comprising: a processor; and a memory device with stored
instructions that, when executed by the processor, cause the system
to: estimate a rate or a frequency of a quasi-periodic signal;
define a search window based on the estimated rate or frequency of
the quasi-periodic signal; define a starting position in the
received quasi-periodic signal, the starting position corresponding
to a first maximum; cross-correlate a portion of the quasi-periodic
signal in the search window with a template signal pattern to be
matched to produce a second maximum that is defined by the
controller as a new starting position; and store the new starting
position.
12. The system of claim 11, wherein the memory device further
causes the system to: determine a localized pattern of the template
signal pattern; extract a signal segment from the localized
pattern; analyze the signal segment simultaneously in time and
frequency domains to produce parallel outputs; combine the parallel
outputs via a statistical or mathematical function to produce a
result; and automatically indicate a diagnosis based on the
result.
13. The system of claim 11, wherein the memory device further
causes the system to indicate the diagnosis directly on a device
selected from a group consisting of the electronic stethoscope, a
second device attached to the electronic stethoscope, and a third
device in wireless communication with the electronic
stethoscope.
14. The system of claim 13, wherein the memory device further
causes the system to indicate the diagnosis via one or more of a
display and audio output of a portable device.
15. The system of claim 13, wherein the memory device further
causes the system to: receive a representation of the
quasi-periodic signal; remove a DC component from the received
signal to produce a purely time-varying signal; filter the
time-varying signal to produce a pre-processed signal;
auto-correlate at least a portion of a representation of the
pre-processed signal with itself and storing in a memory device a
corresponding auto-correlation output for the at least portion of
the representation of the pre-processed signal; apply a biphasic
tapering function to the auto-correlation output, the tapering
function including a time constant parameter that is a function of
the quasi-periodic signal and producing the first maximum; and
store a representation, based on the first maximum, as an
indication of a rate or a frequency of the quasi-periodic
signal.
16. The system of claim 11, wherein the memory device further
causes the system to: save the representation in a file of a
predetermined file format; send the file in an e-mail; print the
file; and store the file on a portable device.
17. The system of claim 16, wherein the memory device further
causes the system to: receive the quasi-periodic signal as an
auscultation signal; record raw auscultation data associated with
the auscultation signal; send the raw data auscultation data to a
hospital information system; analyze, via computation services of
the hospital information system, the raw auscultation data; and
store, on a memory device of the hospital information system,
analysis results for later review by a medical professional.
18. The system of claim 17, wherein the memory device further
causes the system to: retrieve a patient list from a hospital
information system; select a patient from the patient list; obtain
patient specific parameters; and transfer to the hospital
information system, via a data connection, one or more of patient
data, raw auscultation data, the patient specific parameters, and a
diagnosis suggestion.
19. The system of claim 17, wherein the memory device further
causes the system to: receive the quasi-periodic signal as an
auscultation signal; record auscultation data associated with the
auscultation signal; provide user setting options for modifying
settings of the auscultation device; communicate between a medical
professional and an user of the auscultation device; alert the
medical professional when the auscultation data is uploaded and
available for access; allow the medical professional to access the
auscultation data.
20. The system of claim 11, wherein the memory device further
causes the system to: send raw data of the rate or the frequency of
the quasi-periodic signal to a hospital information system;
analyze, using another controller of the hospital information
system, the raw data; and store, on a memory device of the hospital
information system, results of the analysis for later review by a
medical professional.
Description
CROSS-REFERENCE TO RELATED APPLICATIONS
[0001] This application claims the benefit of and priority to U.S.
Provisional Patent Application No. 61/787,998, titled "Automated
Diagnosis-Assisting Medical Devices Utilizing Rate/Frequency
Estimation And Pattern Localization Of Quasi-Periodic Signals" and
filed on Mar. 15, 2013, which is incorporated herein by reference
in its respective entirety.
COPYRIGHT
[0002] A portion of the disclosure of this patent document contains
material which is subject to copyright protection. The copyright
owner has no objection to the facsimile reproduction by anyone of
the patent disclosure, as it appears in the Patent and Trademark
Office patent files or records, but otherwise reserves all
copyright rights whatsoever.
FIELD OF THE INVENTION
[0003] Various aspects of the present disclosure relate to the
estimation of the rate or frequency and to the localization of
similar patterns in quasi-periodic signals. More specifically, for
example, the signals are not limited to being quasi-periodic and
are often overlaid with noise or other artifacts. More generally,
some aspects relate to automated signal processing of sounds
originating from various body structures for providing clinical
referral conditions at a site, such as the patient's site.
BACKGROUND
[0004] Analyzing quasi-periodic signals is very common, e.g.,
analyzing sounds originating from the heart or lungs, and has long
been a tool for evaluating conditions of subjects or patients.
Since the existence of the stethoscope, the electro-cardiogram
device, and similar devices, such practices have been formalized.
In the case of the stethoscope, for example, the sound is detected
non-invasively at the surface of the skin and evaluated by a
skilled practitioner. This is a standard screening method performed
worldwide and called auscultation. Interestingly, auscultation is
also one of the few remaining routine medical procedures in which
the diagnosis is made purely by the medical professional who
listens and interprets the sounds originating from the heart or
lungs based solely on his or her training and experience.
[0005] With no objective and comparable means of evaluation, the
quality of the subjective human diagnosis is solely dependent on
the investigating medical professional, inevitably leading to a
lack of objectivity and a high variability of findings between
medical professionals. As such, making a diagnosis is vulnerable to
human error and subjectivity due to a multitude of potential causes
(e.g., stressful working environment, lack of sleep, etc.)
associated with the medical profession. All of the above identified
problems create a huge burden of responsibility for the medical
professionals. Additionally, without an independent system in place
that supports and documents the medical professionals' findings
from the auscultation (including objective, clinically tested,
investigator-independent, patient-specific parameters), the medical
professionals lack an objective basis to defend their subjective
findings.
[0006] Since the rise of high-speed computing, increasing attention
has focused on analyzing digitized quasi-periodic signals through
digital signal processing ("DSP") techniques. Often, the DSP
techniques have been used simply for determining the rate,
frequency, or steadiness of such signals. More recently, the DSP
technique have also been used to determine pathological conditions
of a medical subject.
[0007] The problem or challenge with such analyses lies in the
reliable extraction and classification of significant features,
often hidden in the recordings of such variable biological signals.
This leads to the importance of proper or correct signal
segmentation, which is often performed manually or sometimes
automatically on good quality signals. In reality, recorded
biological signals in a clinical environment are not of "good"
quality, in which case existing systems struggle to yield reliable
and robust results. For example, most current approaches of rate
detection are simply triggered by the presence of a certain energy
level in the signal, which is problematic in environments
containing other noise or sounds, such as, for example, people
walking or talking, other machines, traffic noise, etc. Other
approaches are also detrimental because they require external
input, such as, for example, electro cardiogram data, to achieve
correct signal segmentation.
[0008] Therefore, there is a need, for example, for handling
uncorrelated noise and variations in the periodicity of such
variable input signals, and/or for estimating signal rates or
frequencies, as well as recognizing and identifying locations of
similar patterns. For example, such technology can be utilized as a
standalone device or as part of a larger system for automated
diagnoses of quasi-periodic signals.
SUMMARY OF THE INVENTION
[0009] According to an aspect of the present disclosure, a method
or algorithm is disclosed that estimates a rate or frequency of
quasi-periodic signals and localizes similar patterns in
quasi-periodic signals without requiring high-end computing powers.
Quasi-periodic signals are signals that can potentially be highly
irregular while still containing some repeating, often hidden,
features. Exemplary signals include biological signals that are
concealed by noise and artifacts from various sources, such as
originating internally (from a body structure) or externally (from
the environment), and that are independent of the signal type or
acquisition (e.g., electrical, mechanical, optical, acoustical,
etc.). The signals are typically, but not necessarily limited to
being quasi-periodic, and are often overlaid with noise or other
artifacts.
[0010] Another aspect of the present disclosure is directed to
determining a representative estimation of the rate of such
signals, e.g., heart beat frequency or breathing frequency. The
signal rate from biological sources is often a parameter of
interest in clinical settings, but can also be utilized in
subsequent or related signal processing stages to perforin further
analysis. The algorithm includes utilizing a combination of
auto-correlation functions, tapering functions, and/or progressive
signal splitting and statistical tools to analyze quasi-periodic
signals.
[0011] Furthermore, yet another aspect of the present disclosure
utilizes signal templates, e.g. a single representative heartbeat
in a series of heartbeats or an analytical signal that shows
similar features as the pattern of interest in the target signal,
to search throughout the entire signal for locations where similar
signal patters (or shapes) are found. The resulting locations are
stored and returned. The algorithm utilizes a sequence of
cross-correlation and windowing functions in combination with
signal rate estimation that makes the algorithm robust against
changes in periodicity, noise, and artefacts.
[0012] According to yet another aspect of the present disclosure, a
method or algorithm is disclosed that changes traditional functions
of electronic stethoscopes from a device typically capable of
recording, storing, and manipulating data, into a device that
automatically delivers diagnostic results for clinically relevant
referral conditions directly to patient's site. By utilizing, for
example, parallel system processing, involving novel algorithms,
and physiological parameters that are optimized with findings from
clinical studies, an embodiment of the present disclosure relates
to a method of analyzing and diagnosing digital physiological
signals that were recorded with commercially available electronic
stethoscopes.
[0013] According to yet another aspect of the present disclosure, a
method of estimating a frequency of a quasi-periodic signal is
performed directly by an electronic stethoscope as illustrated in
FIG. 3 and FIG. 4. Alternatively, the method is performed by a
small, external and portable device connected or linked to the
stethoscope, as illustrated in FIG. 5 (e.g., a small device,
tablet, smartphone, etc.), but that does not require high-end
computing powers. The utilization of one or more aspects of the
disclosed algorithm using standard computing resources, e.g., found
in state-of-the-art smartphones or tablet computers, is possible
due to various attributes of the algorithm, such as further
described below.
[0014] According to one attribute, the algorithm uses methods like
auto-correlation or cross-correlation, which can be computed very
efficiently by using time-frequency conversion to perform such
operations. Microprocessors often provide optimized implementations
of such time-frequency conversions, such as Fast Fourier Transform
("FFT"), and, therefore, significantly boosting time-domain
operations.
[0015] According to another attribute, the algorithm enables fast
and efficient computation by using pre-trained classifiers (e.g.,
neural networks, support vector machines, Bayesian networks, etc.).
The pre-trained classifiers facilitate new data to be classified
with simple and computationally efficient operations (e.g., matrix
multiplications). For this approach, parameters are determined with
training data. For example, in reference to neural networks,
weights and biases determined with training data. Or, in reference
to support vector machines, the location of the support vectors in
the hyperspace is determined with training data. Comprehensive and
well classified training data is useful for a good pre-training of
classifiers. The training data of the disclosed algorithm includes,
for example, raw phonocardiogram data and/or corresponding
diagnoses (obtained using, for example, echocardiography as the
gold standard method for diagnosing heart defects). With a
comprehensive training set, a classifier can be optimized (or,
pre-trained) and applied to new data, which enables fast and
efficient computations. In contrast, so-called lazy-learning
methods use the whole available data set (stored locally) and
compare new data against the whole training set for classification.
The lazy-learning methods lead to higher space requirements for
storing the training data set and/or to increased computational
costs for performing the classification.
[0016] According to yet another attribute, the features of interest
(or inputs to the classifier) are determined in advance by feature
selection algorithms (e.g., sequential floating forward selection),
which reduce feature space. Features are also analyzed using
statistical tools such as a principal component analysis, which
results in linearly uncorrelated variables and which further
optimizes the feature space. Hence, only the most powerful and
meaningful features are selected for the algorithm, increasing its
computational efficiency and robustness against noise and
outliers.
[0017] According to yet another aspect of the present disclosure, a
method and/or system includes combining an electronic stethoscope
with a portable device (FIG. 5) for automated analysis and
diagnosis-support for stethoscope-based auscultation. The method
and/or system utilizes one or more of the algorithms described
below in reference to FIGS. 1, 2, and/or 6. The analysis is
performed by the portable device, which provides results including
a set of patient-specific parameters and/or indicators. The results
are investigator independent and include medical and technical
parameters, such as heart and/or breathing rate, heart and/or
breathing rate variability, systolic and diastolic energy, signal
curve, diagnosis suggestion, etc. At least some of these objective
parameters and/or other results are displayed and/or stored on the
portable device as a means for diagnosis support for the medical
professional or other user.
[0018] According to yet another aspect of the present disclosure, a
bidirectional system architecture is illustrated in FIG. 7 for
enabling a device to be utilized for one or more of the following
purposes: [0019] (i) documentation purposes including, for example,
saving all data and results in a common file format (e.g., PDF
format), printing, emailing, bidirectional integration into a
hospital information system ("HIS"), and/or efficient filing of all
data and results to a patient's medical file; [0020] (ii) teaching,
training, research, and/or presentation purposes by wirelessly
connecting the portable device to a single or multiple other
portable devices that receive all data, including the results
obtained with the utilization of one or more of the described
algorithms; and/or [0021] (iii) bidirectional tele-auscultation for
remote auscultation allowing the user to remotely control settings
of an electronic stethoscope at a patient's site (e.g., change
filters, adjust volumes, etc.), to communicate with a person
operating the electronic stethoscope (e.g., instructing the person
to change the position of the stethoscope), and to further provide
documentation and HIS integration.
[0022] According to yet another aspect of the present disclosure, a
bidirectional system architecture is disclosed as illustrated in
FIG. 7 and in which the HIS is utilized to host or run one or more
of the described algorithms. The system allows a user to access the
system via a portable device or any computer connected to the HIS.
Optionally, recorded signal data is uploaded and/or stored in the
HIS. The data is analyzed by the HIS directly and/or the data is
downloaded onto the portable device for later or remote
analysis.
[0023] According to other aspects of the present disclosure, the
device does not require any external input from medical
professionals or other devices (e.g., ECG), does not require
traditional auscultation techniques to be modified, does not
require especially quiet environments (such as, for example, the
holding of breath by the patient during auscultation), and/or does
not require manual or semi-automated analysis by a medical
professional. Alternatively, adding an external input by the user
is optional and does not hinder the device from performing its
tasks. In fact, the external input might potentially even increase
the accuracy of the results.
[0024] By way of example, in reference to a phonocardiogram
analysis, the age of the patient is a helpful parameter for
narrowing down the range of likely heart rates and possible
diseases (e.g., a specific classification of a heart murmur). A
newborn, for example, usually has a heart rate greater than 100
beats per minute and the range of possible diseases is generally
different than, for example, for a child greater than 2 years of
age. One or more features of the present disclosure are beneficial
to and enhance existing electronic stethoscopes by increasing their
function as a medical device and informing medical staff within a
short period of time whether physiological signals are healthy or
require further medical attention. Thus, one or more features of
the present disclosure can be utilized directly on an electronic
stethoscope or in combination with an electronic stethoscope and a
portable device, wherein computations and interaction (e.g.,
visualization of the findings) with medical staff are achieved
through the portable device.
[0025] According to one embodiment of the present disclosure, a
method is directed to processing a quasi-periodic signal and
includes receiving, using a controller, a representation of a
quasi-periodic signal, and removing, using the controller, a DC
component from the received signal to produce a purely time-varying
signal. The time-varying signal is filtered, using the controller,
to produce a pre-processed signal, and at least a portion of a
representation of the pre-processed signal is auto-correlated,
using the controller, with itself. A corresponding auto-correlation
output is stored in a memory device for the at least portion of the
representation of the pre-processed signal. A biphasic tapering
function is applied, using the controller, to the auto-correlation
output, the tapering function including a time constant parameter
that is a function of the quasi-periodic signal and producing a
first maximum. A representation, based on the first maximum, is
stored in the memory device as an indication of a rate or a
frequency of the quasi-periodic signal.
[0026] According to another embodiment of the present disclosure, a
system for processing a quasi-periodic signal includes a processor
and a memory device stored with instructions that, when executed by
the processor, cause the system to receive a representation of the
quasi-periodic signal. A DC component is removed from the received
signal to produce a purely time-varying signal. The time-varying
signal is filtered to produce a pre-processed signal and at least a
portion of a representation of the pre-processed signal is
auto-correlated with itself. A biphasic tapering function is
applied to the auto-correlation output, the tapering function
including a time constant parameter that is a function of the
quasi-periodic signal and producing a first maximum. A
representation, based on the first maximum, is stored in the memory
device as an indication of a rate or a frequency of the
quasi-periodic signal.
[0027] According to yet another embodiment of the present
disclosure, a method is directed to localizing a pattern in a
quasi-periodic signal and includes estimating, using a controller,
a rate or a frequency of a quasi-periodic signal. A search window
is defined, using the controller, based on the estimated rate or
frequency of the quasi-periodic signal. A starting position is
defined, using the controller, in the received quasi-periodic
signal, the starting position corresponding to a first maximum. A
portion of the quasi-periodic signal in the search window is
cross-correlated, using the controller, with a template signal
pattern to be matched to produce a second maximum that is defined
by the controller as a new starting position. The new starting
position is stored using the controller.
[0028] According to yet another embodiment of the present
disclosure, a system is directed to localizing a pattern in a
quasi-periodic signal and includes a processor and a memory device.
The memory device has stored instructions that, when executed by
the processor, cause the system to estimate a rate or a frequency
of a quasi-periodic signal and define a search window based on the
estimated rate or frequency of the quasi-periodic signal. A
starting position is defined in the received quasi-periodic signal,
the starting position corresponding to a first maximum. A portion
of the quasi-periodic signal in the search window is
cross-correlated with a template signal pattern to be matched to
produce a second maximum that is defined by the controller as a new
starting position. The new starting position is stored.
[0029] According to yet another embodiment of the present
disclosure, a system is directed to processing a quasi-periodic
signal and includes an electronic stethoscope for recording a
quasi-periodic signal, a processor, and a memory device with stored
instructions. When executed by the processors, the stored
instructions cause the system to receive a representation of the
quasi-periodic signal, and to remove a DC component from the
received signal to produce a purely time-varying signal. The
time-varying signal is filtered to produce a pre-processed signal,
and at least a portion of a representation of the pre-processed
signal is auto-correlated with itself. A corresponding
auto-correlation output is stored in a memory device for the at
least portion of the representation of the pre-processed signal. A
biphasic tapering function is applied to the auto-correlation
output, the tapering function including a time constant parameter
that is a function of the quasi-periodic signal. The tapering
function further produces a first maximum. A representation, based
on the first maximum, is stored in the memory device as an
indication of a rate or a frequency of the quasi-periodic
signal.
[0030] The foregoing and additional aspects and embodiments of the
present disclosure will be apparent to those of ordinary skill in
the art in view of the detailed description of various embodiments
and/or aspects, which is made with reference to the drawings, a
brief description of which is provided next.
BRIEF DESCRIPTION OF THE DRAWINGS
[0031] These and other features of exemplary implementations of the
present disclosure will become apparent from the description, and
the accompanying drawings. According to common practice, various
features of the drawings are not to scale but are purposefully
modified arbitrarily for improved clarity wherein:
[0032] FIG. 1 is a diagrammatic illustrating a process of rate or
frequency estimation in quasi-periodic signals.
[0033] FIG. 2 is a flowchart illustrating a process for
localization of similar patterns in quasi-periodic signals.
[0034] FIG. 3 is a representative illustration of an electronic
stethoscope that includes a display for visual indication of a
resulting diagnosis.
[0035] FIG. 4 is a representative illustration of an electronic
stethoscope that acoustically indicates a resulting diagnosis.
[0036] FIG. 5 is a representative illustration of an electronic
stethoscope that is connected to an external, portable device
(e.g., a small device, tablet, smartphone, etc.) for indication
(visually, acoustically, or otherwise) of resulting findings
including diagnosis suggestion.
[0037] FIG. 6 is a flowchart illustrating an exemplary method of
digital signal processing of physiological signals.
[0038] FIG. 7 outlines a bidirectional system architecture for
achieving documentation, teaching, and/or bidirectional
tele-auscultation.
DETAILED DESCRIPTION OF THE INVENTION
[0039] Referring to FIG. 1, a diagrammatic illustrates a
rate/frequency estimation algorithm in accordance with one aspect
of the present disclosure. At 101, a quasi-periodic input signal,
such as an acoustical signal indicative of a physiological rhythm
(e.g., heartbeat, respiration), is loaded. At 102, a DC component
is removed from the input signal, s, according to
s.sub.DCrem=s-mean(s), where
mean ( x ) = 1 N n = 1 N x ( n ) ##EQU00001##
where is the mean operator, s.sub.DCram, is the input signal having
its DC component removed, and N is the length of x.
[0040] At 103, filtering of the input signal is applied to produce
a pre-processed signal that emphasizes the quasi-periodic patterns
of the signal for rate estimation (e.g., the heart sound S1 and S2
in a phonocardiogram). The filtering is performed with a standard
band-pass filter (high-pass filtering and/or low-pass filtering) or
with wavelet filtering. In accordance with wavelet filtering, the
signal is decomposed into detail and approximation coefficients,
and, as such, thresholding of the detail coefficients with
subsequent reconstruction of the signal enables noise removal.
Corresponding cut-off frequencies, filter types, and threshold
levels in 103 are dependent on the type of input signal. Examples
of input signals include heart sounds, breathing sounds, bowel
sounds, quasi-periodic signals, etc. Furthermore, threshold levels
are not directly dependent on the type of input signal, but are
computed specifically for each input signal.
[0041] At 104, from the now pre-processed signal, signal energy is
calculated and normalized, e.g.,
s norm = s filt / 1 N n = 1 N s filt ( n ) 2 , ##EQU00002##
where the denominator corresponds to the root mean square of the
signal, where s.sub.filt is the filtered, pre-processed signal, and
N denotes the length of s.sub.filt. If permitted by signal length,
at 105 the signal energy is split into progressively smaller time
domains, by continuously dividing the entire pre-processed signal
energy in halves, thirds, quarters, etc. The splitting of the
signal energy continues as long as the length of the smallest
resulting domain contains sufficient information for a meaningful
analysis.
[0042] For example, if a medical professional is interested in
analyzing heartbeats, the size of the smallest domains would have
to be large enough to cover the main features of a few heartbeats.
Every resulting signal energy domain is stored accordingly in a
memory device of a computer or computing device.
[0043] At 106, for each domain of the energy the auto-correlation
of the domain itself is computed and stored, yielding
auto-correlations for every domain. A normalized version of the
auto- and cross-correlation is used, which compensates for the
differences in signal magnitudes and properly correlates a shorter
signal with a longer one. This normalized version divides the
results of the correlation by the energy of the parts of the
signals that were effectively correlated.
[0044] First, the shorter signal s.sub.2 is zero-padded to have
equal length as the longer signal s.sub.1. Second, the standard
cross-correlation of s.sub.1 and s.sub.2 is performed by
temp=s.sub.1*s.sub.2, where temp includes only the positive terms,
i.e, the second half, of the standard cross-correlation (* is the
cross-correlation operator). Third, the masked energy correlation,
en.sub.m, is computed according to
en.sub.m=s.sub.1.sup.2*ones(length(s.sub.2)), where the latter term
represents a rectangular window with the length equal to the length
of the shorter signal s.sub.2. Fourth, the result of the normalized
cross-correlation, res.sub.cc, is computed according to
res cc = temp abs ( en m ) * abs ( s 2 s 2 ) ) , ##EQU00003##
where the dot product s.sub.2, s.sub.2 is used. Since a convolution
in the time domain corresponds to a multiplication in the frequency
domain, efficient computation is achieved, e.g., computing
auto-correlation of s(t),res.sub.ac=1FFT(F.sub.s(f)*F.sub.s*(f)),
where F.sub.s(f)=FFT(s(t)) is the FFT of s(t), (f) denotes the
complex conjugate, and 1PPT performs the Inverse Fast Fourier
Transform.
[0045] At 107, a tapering function is applied to every
auto-correlation to amplify the relevant maximum in the
auto-correlation. To amplify the first maximum representing the
rate or frequency, the tapering function is biphasic, where the two
phases are selected depending on the input signal but can generally
include a rising edge followed by a trailing edge. The biphasic
function is necessary because the auto-correlation function of
quasi-periodic signals features multiple peaks, and the biphasic
nature of the tapering function allows the selection of the most
probable (single) peak (by tapering other, more improbable peaks)
representing the period of interest. The tapering function
additionally includes a time constant as a parameter that also
depends on the input signal and that is pre-determined from, for
example, values reported in the literature (e.g., average
breathing/heart rates for different patient groups) or suitable
clinical data if available.
[0046] An example of a biphasic tapering function, f.sub.taper(t),
is the following exponential function
f taper ( t ) = t T c - t T c , ##EQU00004##
where T.sub.c is a time constant and t is the time. The two phases
of the tapering function are reflected in the first term, which
represents a linear increase (e.g., a rising edge or phase), and
the second term representing an exponential decline (e.g., a
trailing edge or phase). Although other biphasic tapering functions
could be used to amplify the maximum in the auto-correlation
representing the period time, the exemplary exponential form
described above is suitable for estimating a frequency specific in
phonocardiograms, based on determinations of clinical
phonocardiogram data covered in substantial amount of noise.
[0047] The exemplary biphasic tapering function yielded the best
results for selecting the single peak in the auto-correlation
function that correlates best with the heart rate of the patient.
The maximum of this particular tapering function is at t=.sub.c (as
can be seen by setting the first derivative of
f taper to zero , f taper ' = T c - t T c 2 - t / T c = 0 ) .
##EQU00005##
Furthermore, T.sub.c is computed by t.sub.c=1/f, where f is the
most probable frequency in the signal, which is determined from
reported values in the literature or clinical data. By way of
example, for very young children (during the first few months of
life) a time constant in the range of 0.6 to 0.3 is appropriate and
corresponds to an average heart rate between about 100-200 beats
per minute.
[0048] At 108, the positions of the maxima in the tapered
auto-correlations are computed and stored in a memory device. At
109, standard statistical measures such as, for example, mean,
median, standard deviation, variance, or other tools (e.g., maximum
likelihood estimation) are utilized to determine one representative
position for all maxima. Finally, at 110 the one representative
position of all maxima of the tapered auto-correlations is
converted, yielding the representative signal rate or signal
frequency.
[0049] Referring to FIG. 2, a flowchart is directed to outlining a
process for the localization of similar patterns in quasi-periodic
signals. This process does not require any external input, such as
an ECG signal for segmentation or other purposes. The localization
algorithm calls for a template representing a signal pattern to be
matched to similar patterns throughout the signal. For example, in
a physiological signal of a series of heartbeats, the template can
be one of the heartbeats in the series.
[0050] The template can also be an analytical signal that shows
similar features as the pattern of interest in the target signal.
For example, for a phonocardiogram, a primitive template includes
two waveforms representing S1 and S2 that are shifted in time,
depending on the estimated heart rate. Examples of such waveforms
that feature certain similarities with S1 and S2 include the
Shannon wavelet
.PSI. shannon ( t ) = sin c ( t 2 ) cos ( 3 .pi. t 2 )
##EQU00006##
and the real part of the Morlet wavelet
.PSI. morlet ( t ) = c .sigma. .pi. - 1 4 - t 2 2 ( .sigma. t -
.kappa. .sigma. ) , ##EQU00007##
where c.sub..sigma. and K.sub..sigma. are constants.
[0051] At 201, the template is cross-correlated with the entire
input signal, such as a physiological signal. The maximum of the
cross-correlation represents the best match of the template with
the signal, and at 202 the position of the maximum is defined as
the starting position S1 for the localization algorithm.
[0052] At 203, the localization algorithm checks if the remaining
signal length after the S1 position is long enough to contain the
search window specified at 204. If yes, the algorithm steps to the
right of S1 (forward in time) where a new starting point is defined
for a search window, shown at 204. The step size, as well as the
size of the search window in 204, is based on an estimated signal
frequency or signal rate, which, for example, is estimated with the
algorithm described in FIG. 1 above.
[0053] At 205, the windowed part of the signal is cross-correlated
with the template. At 206, the position of the maximum in the
cross-correlation is computed and stored in a memory device as the
new starting point S.sub.1. The maximum represents the best match
of the template with the signal within the search window. Next, the
localization algorithm goes back to 203 to check if it has arrived
close to the signal end. If not, modules 204-206 are repeated. If
yes, the algorithm goes back to the starting position S1, according
to 207.
[0054] Throughout modules 208-211, the localization algorithm
performs the same operations as in modules 203-206, but instead of
stepping to the right the algorithm keeps stepping to the left of
S1 (back in time). The step size as well as the size of the search
window throughout modules 208-211 are again based on an estimated
signal frequency or signal rate, but are not necessarily the same
as at modules 203-206. Examples of such step sizes suitable for
auscultation data from new-borns are 0.6*T.sub.p for the start and
1.8*T.sub.p for the end of the search
T p = 1 f ##EQU00008##
window to the right, and 1.4*T.sub.p and 0.2*T for the length of
the search window, where is the period.
[0055] When the localization algorithm arrives at a position too
close to the signal start (the left end of the signal), where the
remaining part of the signal to the left is too short to contain a
new search window, 208 ensures that the localization algorithm
jumps to 212, where all stored positions S.sub.i are returned. The
positions S.sub.i represent the locations of the patterns
throughout the signal that match the template. This process can be
repeated for different patterns of the same input signal. The input
signal is not limited to being quasi-periodic and can include a
significant amount of noise or artefacts. Furthermore, this process
is independent of the type of signal or signal acquisition (e.g.,
electrical, mechanical, optical, acoustical, etc.).
[0056] Advantages of using a search window include that the signal
does not have to be strictly periodic and repeating patterns can
still be found. Moreover, the window restricts the search area to a
reasonable size, ensuring that patterns covered in noise can also
be detected using a template containing similar features as the
desired pattern. Ultimately, the lengths and positions of the
search windows depend on the signal, keeping in mind that longer
windows allow the signal to be more irregular while making the
pattern detection in noisy signals more complicated. The position
and length of the window is selected such that it does not contain
two or more of the patterns of interest, otherwise one of the
multiple patterns might be skipped.
[0057] Referring to FIG. 6, a flow chart illustrates modifying an
electronic stethoscope into a diagnosis-assisting tool providing
additional medically relevant information regarding the
physiological signal. At 601, a quasi-periodical, digital
physiological signal is received from an electronic stethoscope.
Then, the signal is pre-processed (e.g. filtered, normalized, etc.)
at 602.
[0058] The beat frequency is estimated at 603 using, e.g., the
algorithm illustrated in FIG. 1 and used in the segmentation stage
at 604, which is also illustrated in FIG. 2. This segmentation at
604 yields the segments of interest of the signal, e.g., the
systole and/or diastole of each heart beat, and/or the inhale
and/or exhale phase of a breath.
[0059] In a parallel system processing stage, for example, at 605,
various analysis modules operating in both the frequency and the
time domain are applied to extract features of the segments
obtained at 604. For example, such features include an Energy
Analysis, Timing Features (e.g., Heart Rate Variability, Duration
of S1/Systole/S2/Diastole, etc.), Fourier Transform, Short Time
Fourier Transform, Higher Order Statistics (e.g., Bispectra,
Gaussian Mixture Models, etc.), Discrete and Continuous Wavelet
Analysis, Fractal Dimension Analysis, Stockwell-Transform, Error
Entropy Analysis, etc. The output of the parallel modules at 605 is
combined at 606 and passed to 607, where the output is used in a
decision making stage. If external data, such as patient age or
certain aspects of patient's medical history, is added,
classification of specific pathologies might yield increasing
accuracy.
[0060] The output of 607 is forwarded to 608 where the diagnosis is
indicated to the user on an electronic display device. The
diagnosis is indicated, for example, via a binary output, via
probabilities, via an acoustic signal, and/or via a visual
interface. The visual interface can include a detailed listing of
the findings, including a diagnosis suggestion, all of which are
optionally stored in a memory device on an electronic stethoscope
or shared, stored or printed through other means as described
above.
[0061] Additionally, still referring to FIG. 6, a basic structure
of a digital signal processing is utilized by one or more aspects
of the invention. Parallelization, in combination with one or more
of the above described algorithms and in combination with specific
parameters obtained through clinical studies enable a fully
automated analysis without requiring any external input in addition
to the physiological signal itself (although external data and/or
parameters can be optionally added). The diagnostic result is
revealed directly on the electronic stethoscope (e.g., FIGS. 3-5)
or the portable device that is connected to the electronic
stethoscope (e.g., on display 502 of FIG. 5), either visually
(through a display 301, 502) or acoustically (e.g., sound emitter
401).
[0062] Some or all modules described above, which have been
described by way of example herein, represent one or more
algorithms that correspond to at least some instructions executed
by one or more controllers to perform the functions or modules
disclosed. Any of the methods or algorithms or functions described
herein can include machine or computer-readable instructions for
execution by a processor, controller, computer, and/or any other
suitable processing or computing device. Any algorithm, software,
and/or method disclosed herein can be embodied as a computer
program product having one or more non-transitory tangible medium
or media, such as, for example, a flash memory, a CD-ROM, a floppy
disk, a hard drive, a digital versatile disk ("DVD)", or other
memory devices. However, persons of ordinary skill in the art will
readily appreciate that the entire algorithm and/or parts thereof
can alternatively be executed by a device other than a controller
and/or embodied in firmware or dedicated hardware (e.g., it can be
implemented by an application specific integrated circuit ("ASIC"),
a programmable logic device ("PLD"), a field programmable logic
device ("FPLD"), discrete logic, etc.).
[0063] Referring to FIG. 5, illustrates a method or system for
automated analysis and diagnosis-support for stethoscope-based
auscultation. An automated analysis and diagnosis-support system
500 includes an electronic stethoscope 501 with signal transmitting
capabilities, which include, for example, an integrated Bluetooth
transmitter and/or an appropriate transmitter for transmitting
signals directly connected (e.g., via an audio jack) to the
electronic stethoscope 501. The transmitter is capable of
converting an analog signal to a digital signal and is optionally
capable of encrypting the signal.
[0064] For example, heart sounds are transmitted to a processing
unit 502, which can be included in a smartphone, a tablet, a
computer, etc. The processing unit 502 automatically analyzes the
transmitted signal with the utilization of one or more of the
algorithms described above. The analysis yields a set of
patient-specific parameters/indicators and results, including
medical and technical parameters such as heart/breathing rate,
heart/breathing rate variability, systolic and diastolic energy,
signal curve, diagnosis suggestion (e.g., through probabilities or
binary output), etc. All or a selection of these objective
parameters and results are displayed and/or stored on the portable
device as a means for diagnosis support for the medical
professional.
[0065] Referring to FIG. 7, a bidirectional system architecture is
illustrated for use with one or more of the algorithms described
above for analyzing a signal, enabling a portable device 703 to be
utilized for (i) documentation, (ii) teaching, and/or (iii)
bidirectional tele-auscultation purposes.
[0066] In reference to documentation, items 701-704 illustrate
utilizing an automated, analysis and diagnosis-support system 500
for documentation purposes. All data and results are saved as a
common file type (e.g., PDF format) on the portable device 703,
printed, and/or emailed. A bidirectional interface between the
portable device 703 and the HIS 704 allows for retrieving of
patient data from the HIS if required by the medical professional.
The bidirectional interface further allows efficient filing of all
data and results to the patient's medical file.
[0067] In reference to teaching, items 701-703 and 705 illustrate
utilizing the automated, analysis and diagnosis-support system 500
for stethoscope-based auscultation as described above. The
teaching, which is optionally directed to achieving training,
research, and/or presentation objectives, is achieved by wirelessly
connecting the portable device 703 to a single or multiple other
portable devices 705. The portable devices 705 receive all data
including the findings of the analysis system. According to one
example, a professor teaches medical students the art of
auscultation by performing auscultation using the electronic
stethoscope 701 on one student and transmitting all related data
and results of the system on the portable device 703 to the
portable devices 705 of other students.
[0068] In reference to bidirectional tele-auscultation, items
701-703 and 706-709 illustrate a possibility for utilizing the
automated, analysis and diagnosis-support system for
stethoscope-based auscultation remotely through bidirectional
tele-auscultation. In such a scenario, the data is transmitted from
the first electronic stethoscope 701 (e.g., operated by a nurse or
by the patient himself), through a data link 702 to the portable
device 703. The data is further transmitted through a data
connection 706 (e.g., the Internet) to a second portable device 707
(operated, e.g., by the medical professional performing the
auscultation), and optionally through another data link 708 to a
second electronic stethoscope 709 (operated, e.g., by the medical
professional performing the auscultation).
[0069] In the above scenario, the first portable device 703
performs the required transmitting functions (with no analysis of
the signal), and the second portable device 709 has the automated,
analysis and diagnosis-support system for stethoscope-based
auscultation running. The bidirectionality of this pathway (701-703
and 706-709) allows the medical professional operating the second
portable device 707 to control the settings of the electronic
stethoscope 701 (e.g., change filters, adjust volumes, etc.) and to
communicate with the person operating the electronic stethoscope
through the first stethoscope 701 or the portable device 703 (e.g.,
instructing the person to change the position of the stethoscope).
Documentation and HIS integration are options in the scenario.
[0070] Optionally, the automated analysis and diagnosis-support
system illustrated in FIG. 5 is installed and/or hosted by the HIS
(e.g., HIS 704 of FIG. 7) and the user accesses the HIS via a
portable device (e.g., portable devices 703, 707 of FIG. 7) or via
a computer connected to the HIS. In such a case, the recorded
signal data is optionally uploaded and/or stored in the HIS and is
analyzed in the HIS directly utilizing one or more of the
algorithms described above, and/or the data is downloaded onto the
portable device for later or remote analysis utilizing one or more
of the algorithms described above.
[0071] By way of a specific example, a medical professional uses an
electronic stethoscope while medically evaluating a 10 year-old
patient. The electronic stethoscope includes a communication input
for connecting with a portable device. The medical professional
connects the portable device (e.g., a smartphone) (a) to the
hospital information system to receive the patient's medical data
(e.g., age, medical history, etc.) and (b) to the electronic
stethoscope. The communication input of the electronic stethoscope
includes, for example, a built-in Bluetooth chip or an external
Bluetooth transmitter connected an audio jack.
[0072] During the medical evaluation in which auscultation is
performed, the electronic stethoscope records an acoustic heart
signal while the medical professional listens to the heart sound.
The electronic stethoscope converts the acoustic heart signal from
an analog format into a digital format and transfers the digitized
signal to the smartphone. The smartphone receives the digitized
signal and processes it, including removing the DC component,
filtering, etc.
[0073] After the processing of the digitized signal, the smartphone
estimates the heart rate by partitioning the signal,
auto-correlating the individual parts, applying a tapering function
to each autocorrelation, and statistically analyzing the maxima of
all autocorrelations. Then, the heart rate serves as the input for
the segmentation stage, where a representative template (e.g., a
template pre-stored on the smartphone) is correlated with the
digitized signal to find the best matches of this template in the
digitized signal within defined search windows.
[0074] The segmentation results from the previous modules are used
in a feature extraction stage, where characteristic properties (or
features) from the digitized signal are extracted (e.g., via
Fourier Transform, Gaussian Mixture Models, Energy Analysis, etc.).
These features serve as the input for a decision stage, in which
the features are classified, for example, via Multilayer
Perceptrons, Support Vector Machines, and/or a combination/cascade
of such classifiers. The classifiers yield a diagnosis suggestion
and/or a set of patient-specific parameters that are displayed for
the medical professional via the smart phone. Results are
optionally stored on the smart phone, and/or printed, sent via
e-mail, and/or stored in the hospital information system.
[0075] Each of these embodiments and obvious variations thereof is
contemplated as falling within the spirit and scope of the claimed
invention, which is set forth in the following claims. Moreover,
the present concepts expressly include any and all combinations and
subcombinations of the preceding elements and aspects.
* * * * *