US-12625037-B2 - System and a computer-implemented method for detecting medical-device errors by analyzing acoustic signals generated by the medical device's components

US12625037B2US 12625037 B2US12625037 B2US 12625037B2US-12625037-B2

Abstract

A system includes a sensor element arranged to detect a spectrum of noise or vibrations of a medical device to be supervised the, a supervising element, whereby the sensor element and the supervising element may communicate with each other and thereby allow for provisioning of data corresponding to the acoustic signals, whereby the system further comprises a localized or distributed detection engine, the detection engine analyzing the data corresponding to the acoustic signals such that typical failures of each of a plurality of individual components of the medical device are distinguished, whereby the system further comprises a notification engine, the notification engine providing indications on the maintenance state of the medical device and/or one or more of the plurality of individual components of the medical device. The disclosure also pertains to a computer-implemented method for determining a maintenance state of a medical device.

Inventors

Aaron Pickering
Ritwika Mukherjee
Felix Adam
Marie Elisabeth Heinrich
Alan Wei Min Tan
Daniel Horcher
Torsten Labs

Assignees

FRESENIUS MEDICAL CARE DEUTSCHLAND GMBH
FRESENIUS MEDICAL CARE HOLDINGS, INC.

Dates

Publication Date: 20260512
Application Date: 20220804

Claims (20)

1 . A system for detecting errors of a medical device by analyzing acoustic or vibration signals generated by at least one component of the medical device, the system comprising: a sensor element arranged to detect a spectrum of noise or vibrations of the medical device and to be arranged at the medical device, a supervising element, whereby the sensor element and the supervising element are adapted to communicate with each other and thereby allow for data corresponding to the acoustic signals to be transmitted between the sensor element and the supervising element, a localized or distributed detection engine, the detection engine adapted to analyze the data corresponding to the acoustic signals such that typical failures of each of a plurality of individual components of the medical device are distinguished, wherein the detection engine is configured to: generate an operation state of the at least one component of the medical device via a first machine-learning model based on the acoustic signals, wherein the operation state is indicative of a type of treatment being performed by the medical device; and determine a maintenance state of the at least one component of the medical device via a second machine-learning model based on the operation state and the acoustic signals; and a notification engine adapted to provide indications on the maintenance state of the medical device or one or more of the plurality of individual components of the medical device.
2 . The system according to claim 1 , whereby the data corresponding to the acoustic or vibration signals is preprocessed to thereby reduce the amount of data corresponding to the acoustic signals.
3 . The system according to claim 1 , whereby the sensor element is remote to the supervising element.
4 . The system according to claim 1 , whereby the detection engine is based on at least one trained neural network.
5 . The system according to claim 1 , whereby the notification engine is adapted to provide a maintenance state via a network interface.
6 . The system according to claim 1 , whereby the detection engine is enabled to self-learn such that newly experienced failures of each of a plurality of individual components may be learned for future detection.
7 . A computer-implemented method for determining a maintenance state of a medical device or components of the medical device, the method comprising: acquiring first acoustic data or first vibration data from a first acoustic signal or a first vibration signal generated by the medical device or the components of the medical device for training a first machine-learning model, wherein the first machine-learning model is configured to generate an operation state of the medical device or components of the medical device based on the first acoustic data or the first vibration data, wherein the operation state is indicative of a type of treatment being performed by the medical device, and wherein a second machine-learning model is configured to determine the maintenance state of the medical device or the components of the medical device based on the operation state and the first acoustic data or the first vibration data; acquiring second acoustic data or second vibration data from a second acoustic signal or a second vibration signal generated by the medical device or the components of the medical device and generating via the first machine-learning model a present operation state of the medical device or the components of the medical device, wherein the present operation state is indicative of a present type of treatment being performed by the medical device; and determining via the second machine-learning model the maintenance state of the medical device or the components of the medical device; wherein the maintenance state is determined at least partly based on (i) the second acoustic data or the second vibration data and (ii) the present operation state generated by the first machine-learning model.
8 . The computer-implemented method according to claim 7 , further comprising providing the maintenance state via a network interface.
9 . The computer-implemented method according to claim 7 , wherein the training is unsupervised.
10 . The computer-implemented method according to claim 7 , wherein the first machine-learning model is configured to determine the maintenance state based on the operation state and the second acoustic data or the second vibration data, and wherein the determination of the maintenance state of the medical device or the components of the medical device is performed via the first machine-learning model.
11 . The computer-implemented method according to claim 7 , wherein acquiring the first acoustic data or first vibration data comprises extracting features from the first acoustic data or first vibration data, wherein the features include at least one of: a root mean square, a spectral skew, and magnitudes.
12 . The computer-implemented method according to claim 7 , wherein acquiring the first acoustic data or first vibration data comprises splitting the first acoustic data or first vibration data into segments to create component-specific fingerprints and failure detection models.
13 . The computer-implemented method according to claim 7 , wherein the first machine-learning model is based on a mean-square error histogram of previous acoustic data or previous vibration data.
14 . The computer-implemented method according to claim 7 , wherein acquiring the second acoustic data or second vibration data comprises transforming the second acoustic data or second vibration data into a time-frequency representation.
15 . A medical device adapted for dialysis treatment, the medical device comprising: a sensor element arranged to detect a spectrum of noise or vibrations of the medical device, a supervising element, whereby the sensor element and the supervising element are adapted to communicate with each other and thereby allow for data corresponding to the acoustic signals to be transmitted between the sensor element and the supervising element, a localized or distributed detection engine, the detection engine adapted to analyze the data corresponding to the acoustic signals such that typical failures of each of a plurality of individual components of the medical device are distinguished, wherein the detection engine is configured to: generate an operation state of the at least one component of the medical device via a first machine-learning model based on the acoustic signals, wherein the operation state is indicative of a type of treatment being performed by the medical device; and determine a maintenance state of the at least one component of the medical device via a second machine-learning model based on the operation state and the acoustic signals; and a notification engine adapted to provide indications on the maintenance state of the medical device or one or more of the plurality of individual components of the medical device.
16 . The medical device according to claim 15 , whereby the data corresponding to the acoustic or vibration signals is preprocessed to thereby reduce the amount of data corresponding to the acoustic signals.
17 . The medical device according to claim 15 , whereby the sensor element is remote to the supervising element.
18 . The medical device according to claim 15 , whereby the detection engine is based on at least one trained neural network.
19 . The medical device according to claim 15 , whereby the notification engine is adapted to provide a maintenance state via a network interface.
20 . The medical device according to claim 15 , whereby the detection engine is enabled to self-learn such that newly experienced failures of each of a plurality of individual components may be learned for future detection.

Description

TECHNICAL FIELD The present disclosure relates to a system and a computer-implemented method for detecting medical-device errors by analyzing acoustic signals generated by the medical device's components. BACKGROUND Within the medical field and in particular within the field of dialysis, a large number of medical devices are used within prolonged periods. Some of these machines may be used by individuals in a home-setting while others may be used in a care-home or in care-centers by one or more individuals. These medical devices may be of a life-supporting nature and sometimes may be used on a regular basis such as for dialysis treatments. Failure of one or more components of the medical device may lead to a life-threatening situation for one or more concerned patients. While newer devices are equipped with a functionality allowing to determine a maintenance status either stand-alone or in connection with a centralized infrastructure, older devices typically do not comprise such functionality. However, there is still a large number of old devices in use. This is to some extent based on the fact that these devices may be used by a user for a long time and also on the fact that the cost of these devices are rather high. Therefore, a need exists to provide such functionality also for older medical devices. Even though some of these medical devices may be furnished with upgrades for their respective hardware and firmware, allowing in cooperation with further elements to share further information, it is to be noted that hardware and firmware upgrades may entail a need for a renewed validation process due to regulatory requirements for the upgraded devices to be used within the medical field. However, such a validation process is both expensive and time consuming. In addition, it is noted that firmware upgrades usually involve high quality requirements in order to be successful and to not render the medical device useless in case of failure. Even then it is to be noted that a firmware upgrade may not provide measured data. That is, while a firmware upgrade may provide data, e.g., relating to the speed of a motor, such information may not be sufficient to determine a type of error in case of a reduced speed. Therefore, much more sophisticated changes of the medical device's hardware implementation would be required. However, such changes are usually not feasible due to lengthy medical-device downtimes and huge additional costs in case of comprehensive hardware changes. SUMMARY The present disclosure proposes a system for detecting errors of a medical device by analyzing acoustic signals generated by a component of the medical device, the system comprising: a sensor element arranged to detect a spectrum of noise or vibrations of a medical device to be supervised and to be arranged at the medical device to be supervised,a supervising element,whereby the sensor element and the supervising element may communicate with each other and thereby allow for provisioning of data corresponding to the acoustic signals,whereby the system further comprises a localized or distributed detection engine, the detection engine analyzing the data corresponding to the acoustic signals such that typical failures of each of a plurality of individual components of the medical device are distinguished,whereby the system further comprises a notification engine, the notification engine providing indications on the maintenance state of the medical device and/or one or more of the plurality of individual components of the medical device. According to some embodiments, the disclosure proposes that the data corresponding to the acoustic signals is preprocessed to reduce the amount of data corresponding to the acoustic signals. According to certain embodiments, the disclosure proposes that the sensor element is remote to the supervising element. According to some embodiments, the disclosure proposes that the detection engine is based on at least one trained neural network. In certain embodiments, the notification engine provides a maintenance state via a network interface. In some embodiments, the detection engine is enabled to self-learning such that newly experienced failures of each of a plurality of individual components may be learned for future detection. The disclosure also proposes a computer-implemented method for determining a maintenance state of a medical device comprising the steps of: acquiring first acoustic data from an acoustic signal generated by the medical device and/or the medical device's components for (unsupervised/supervised) training a first and/or second machine-learning model,wherein the first machine-learning model is configured to generate an operation state of the medical device based on the acoustic data andwherein the first or second machine-learning model is configured to determine a maintenance state based on the operation state and the acoustic data.acquiring second acoustic data from an acoustic signal generated