# US 11,307,924 B2 · Sequence mining in medical IoT data

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

- Number: US 11,307,924 B2
- Document ID: US11307924B2
- Status: Granted
- Kind: B2
- Application number: 16/819,939
- Filed: 2020-03-16
- Published: 2020-10-01
- Granted: 2022-04-19
- Earliest claimed: 2019-03-25
- Assignees: Siemens Healthcare GmbH
- Applicants: Siemens Healthcare GmbH
- Inventors: Tobias Hipp; Thomas Hubauer; Ulli Waltinger
- Web page: https://patentgrep.com/patent/US11307924B2

## CPC

- G06F 11/079: {Root cause analysis, i.e. error or fault diagnosis (in a hardware test environment G06F11/22; in a software test environment G06F11/36)}
- G16H 40/40: for the management of medical equipment or devices, e.g. scheduling maintenance or upgrades
- G05B 23/0283: {Predictive maintenance, e.g. involving the monitoring of a system and, based on the monitoring results, taking decisions on the maintenance schedule of the monitored system; Estimating remaining useful life [RUL] (preventive maintenance, i.e. planning maintenance according to the available resources without monitoring the system G06Q10/06)}
- G06F 11/0736: {in functional embedded systems, i.e. in a data processing system designed as a combination of hardware and software dedicated to performing a certain function (testing or monitoring of automated control systems G05B23/02)}
- G06F 11/0751: {Error or fault detection not based on redundancy (power supply failures G06F1/30; network fault management H04L41/06)}
- G06F 11/0778: {Dumping, i.e. gathering error/state information after a fault for later diagnosis}
- G06N 5/04: Inference or reasoning models
- G06Q 10/0631: Resource planning, allocation, distributing or scheduling for enterprises or organisations
- G06Q 10/20: Administration of product repair or maintenance
- G05B 2219/45169: Medical, rontgen, x ray
- G16H 50/70: for mining of medical data, e.g. analysing previous cases of other patients
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)
- G16H: HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
- G05B: CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS (systems for controlling or regulating non-electric variables G05D; systems for regulating electric or magnetic variables G05F; control devices or systems insofar as characterised by mechanical features only G05G)
- G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06Q: INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR

## Abstract

A computer-implemented method for configuring a computing device for predictive maintenance, a computer-implemented method for predictive maintenance as well as a predictive maintenance apparatus are disclosed. Training log files including event sequences are examined iteratively for sequences of increasing length in order to determine a set of configuration data containing event sequences that have high predictive power for a system failure. Forward and backward gap values are defined such that not only sequences in the exact same temporal order as in the training log files are examined but also sequences with slightly different temporal ordering. In this way, possibly imprecise and/or incorrect time stamps in log files are compensated.

## Claims

### Claim 1

A computer-implemented method for maintenance of a medical imaging device, the method comprising:
providing at least one training log file including a list of events, each of the at least one training log file corresponding to at least one medical imaging device, and each event in the list of events including at least a unique identifier and a timestamp;
applying a SPADE algorithm to the at least one training log file to determine, as a configuration dataset, a sub-set including sequences of events that fulfill a finishing criterion;
configuring a computing device to recognize a sequence of events, in the configuration dataset as indicative of a failure of the medical imaging device, in an event sequence list provided to the computing device;
providing at least a first event sequence list to the computing device, the first event sequence list including a first list of events, and the first event sequence list corresponding to the medical imaging device;
predicting, using the computing device, a failure of the medical imaging device based on the first event sequence list; and
at least one of (i) automatically scheduling maintenance of the medical imaging device or (ii) rescheduling or re-tasking at least one function of the medical imaging device, in response to predicting the failure of the medical imaging device, wherein
the re-tasking re-tasks the at least one function of the medical imaging device to a different medical imaging device, and
within the SPADE algorithm, each sub-set of events for a same medical imaging device is considered to be a sequence, or part of a sequence, when an absolute value of a difference between a first event timestamp and a second event timestamp, of the sequence of events, is less than a defined backward gap value, and when an absolute value of a difference between a third event timestamp and the first event timestamp, of the sequence of events, is less than a defined forward gap value.

### Claim 2 (depends on 1)

The method of claim 1, wherein the applying a SPADE algorithm comprises:
labelling, in each respective training log file of the at least one training log file, (i) all events within at least one of a defined timespan and a defined number of events before any system failure of the medical imaging device corresponding to the respective training log file occurred with preliminarily positive indicators, and (ii) all other events with preliminarily negative indicators, and
providing the at least one training log file as input to a first iteration, i=1;
wherein the first iteration, i=1, includes
examining sequences within the input provided to the first iteration, i=1, with a first sequence length l1=1,
selecting, from the sequences, a first sub-set of sequences according to a first selecting criterion for the first iteration, i=1, wherein the first selecting criterion is based at least on at least one of the preliminarily positive indicators or the preliminarily negative indicators, and
providing the first sub-set of sequences to an i-th iteration, with i>1, and
wherein the i-th-iteration includes
examining possible sequences starting with an (i−1)-th sub-set of sequences, which is provided to the i-th iteration, with a sequence length li, wherein for all i, li≤li−1,
selecting, from the possible sequences, an i-th sub-set of possible sequences according to an i-th selecting criterion for the i-th iteration, wherein the i-th selecting criterion is based at least on at least one of the preliminarily positive indicators or the preliminarily negative indicators,
determining whether the i-th subset fulfils a finishing criterion for the i-th iteration,
providing the i-th subset as a configuration dataset in response to determining that the i-th subset fulfils the finishing criterion for the i-th iteration; and
providing the i-th subset to an (i+1)-th iteration and performing the (i+1)-th iteration in response to determining indicating that the i-th subset does not fulfil the finishing criterion for the i-th iteration.

### Claim 3 (depends on 1)

The method of claim 1, wherein the defined backward gap value is less than the defined forward gap value.

### Claim 4 (depends on 3)

The method of claim 3, wherein at least one of
the defined backward gap value is between 0.1 seconds and 1.5 seconds, or
the defined forward gap value is between 1 second and 5 seconds.

### Claim 5 (depends on 2)

The method of claim 2, wherein finishing criterion of all iterations includes a criterion that subsets selected to be provided as the configuration dataset, must have a defined target sequence length.

### Claim 6 (depends on 2)

The method of claim 2, wherein log files of a plurality of medical imaging devices are collected at a data warehouse, and wherein the log files collected are split, according to a defined ratio, into training data and testing data, and wherein the training data are provided as the at least one training log file for the labelling and as input into the first iteration.

### Claim 7 (depends on 2)

The method of claim 2, further comprising:
calculating, for each of the sequences examined in each iteration, a positive predictive value (PPV) score, the PPV score indicating a percentage of cases in which a corresponding sequence of events has correctly indicated a system failure; and
the i-th selecting criterion includes a criterion that the PPV score is greater than a threshold value for each iteration and that a minimum number of examined sequences is to be selected for each iteration i.

### Claim 8 (depends on 7)

The method of claim 7, wherein the threshold value is adjusted from the i-th iteration to an (i+1)-th iteration.

### Claim 9 (depends on 8)

The method of claim 8, wherein the threshold value of the i-th iteration is less than or equal to the threshold value of the (i+1)-th iteration.

### Claim 10 (depends on 7)

The method of claim 7, wherein the minimum number of examined sequences to be selected for each iteration is less than or equal to a minimum number for the (i+1)-th iteration.

### Claim 11 (depends on 2)

The method of claim 2, wherein the i-th selecting criterion includes a criterion that only examined sequences, which occur on a defined minimum number of different medical imaging devices according to a corresponding at least one training log file, are selectable.

### Claim 12 (depends on 2)

The method of claim 2, wherein the i-th selecting criterion includes a criterion that only examined sequences, which occur on a defined minimum number of different days according to a corresponding at least one training log file, are selectable.

### Claim 13 (depends on 1)

The method of claim 1, wherein the configuration dataset is at least one of cleansed or partially merged using a clustering method.

### Claim 14 (depends on 1)

A predictive maintenance apparatus, comprising:
an input interface configured to receive at least one event sequence list, each at least one event sequence list including a list of events, and each at least one event sequence list corresponding to at least one medical imaging device, each event of the list of events including at least a unique identifier and a timestamp;
a computing device configured according to the method of claim 1;
and
an output interface configured to generate a maintenance signal indicating the predicted failure of the medical imaging device.

### Claim 15 (depends on 1)

A computer program product including a non-transitory computer-readable medium storing instructions that, when executed by a predictive maintenance apparatus or a computing device, cause the predictive maintenance apparatus or the computing device to carry out the method of claim 1.

### Claim 16 (depends on 1)

A non-transitory computer readable medium storing a computer program including instructions that, when executed by a predictive maintenance apparatus or a computing device, cause the predictive maintenance apparatus or the computing device to carry out the method of claim 1.

### Claim 17 (depends on 3)

A non-transitory computer readable medium storing a computer program including instructions that, when executed by a predictive maintenance apparatus or a computing device, cause the predictive maintenance apparatus or the computing device to carryout the method of claim 3.