# US 2023/0411005 A1 · Predicting surgery duration

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

- Number: US 2023/0411005 A1
- Document ID: US20230411005A1
- Status: Published application
- Kind: A1
- Application number: 18/209,522
- Filed: 2023-06-14
- Published: 2023-12-21
- Earliest claimed: 2022-06-17
- Assignees: Koninklijke Philips N.V.
- Applicants: Koninklijke Philips N.V.
- Inventors: Mariana Nikolova-Simons; Marko Djikic; Rikkert Keldermann; Leon Montenij
- Web page: https://patentgrep.com/patent/US20230411005A1

## CPC

- G16H 50/20: for computer-aided diagnosis, e.g. based on medical expert systems
- G06N 20/20: Ensemble learning
- G16H 20/40: relating to mechanical, radiation or invasive therapies, e.g. surgery, laser therapy, dialysis or acupuncture
- G16H 50/70: for mining of medical data, e.g. analysing previous cases of other patients
- G16H 70/20: relating to practices or guidelines
- G16H 10/60: for patient-specific data, e.g. for electronic patient records
- G16H 40/00: ICT specially adapted for the management or administration of healthcare resources or facilities; ICT specially adapted for the management or operation of medical equipment or devices
- G16H 40/20: for the management or administration of healthcare resources or facilities, e.g. managing hospital staff or surgery rooms
- G16H: HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
- G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS

## Abstract

A system and method are provided for generating a predictive model for predicting a surgery duration. The system and method use a feature selection technique to identify a set of features in training data, which set of features is predictive of the surgery duration, train a number of predictive models using the set of features as input and the surgery duration as prediction target, wherein the predictive models include at least a linear predictive model and a non-linear predictive model, and generate an ensemble model which combines at least two of the predictive models. Such an ensemble model may optimally combine linear and non-linear predictions and therefore allow linear and non-linear relationships between features and the surgery duration to be taken into account. Advantageously, more accurate surgery planning may safeguard the health of patients, for example by ensuring that there are sufficient resources available for acute surgeries, or by avoiding that elective surgeries have to be postponed due to a presumed lack of resources.

## Claims

### Claim 1

A system for generating a predictive model for predicting a surgery duration, comprising:
an input interface for accessing medical data, the medical data comprising records of surgeries, wherein a record is indicative of at least a type of surgery and a surgery duration;
a processor subsystem configured to generate a predictive model for predicting the surgery duration by:
using a feature selection technique to identify a set of features in training data, which set of features is predictive of the surgery duration, wherein the training data comprises a first part of the medical data;
training a number of predictive models using the set of features in the training data as input and the surgery duration as prediction target, wherein the predictive models include at least a linear predictive model and a non-linear predictive model;
using a second part of the medical data, evaluating a performance of each the predictive models in predicting the surgery duration, wherein the evaluating of the performance comprises using a performance metric which characterises a time difference between a predicted surgery duration and an actual surgery duration;
based on the performance of the predictive models, generating an ensemble model which combines at least two of the predictive models; and
outputting the ensemble model for use in predicting the surgery duration.

### Claim 2 (depends on 1)

The system according to claim 1, wherein the performance metric characterises whether the time difference between the predicted surgery duration and the actual surgery duration is positive or negative and a degree of the difference.

### Claim 3 (depends on 2)

The system according to claim 2, wherein the performance metric categorizes the time difference into a number of categories, wherein the categories include at least a first category indicating that the predicted surgery duration exceeds the actual surgery duration by a first margin and a second category which indicates that the predicted surgery duration is less than the actual surgery duration by a second margin.

### Claim 4 (depends on 1)

The system according to claim 1, wherein in evaluating the performance, a positive time difference between the predicted surgery duration and the actual surgery duration is weighted differently than a negative time difference.

### Claim 5 (depends on 1)

The system according to claim 1, further comprising a user interface subsystem comprising a user input interface to a user input device for receiving user input and a display output to a display for displaying output of the system, wherein the processor subsystem is configured to, using the user interface subsystem:
enable a user to evaluate the performance of a previous predictive model for predicting the surgery duration using the performance metric;
request a new predictive model to be generated by the system if the performance of the previous predictive model is deemed insufficient by the user or the system.

### Claim 6 (depends on 5)

The system according to claim 5, wherein the processor subsystem is configured to, using the user interface subsystem, enable the user to further evaluate the performance of the new predictive model using the performance metric.

### Claim 7 (depends on 1)

The system according to claim 1, wherein the processor subsystem is configured to:
generate two or more ensemble models which each combines at least two of the predictive models;
using a third part of the medical data, evaluating a performance of the two or more ensemble models in predicting the surgery duration, wherein the evaluating of the performance comprises using the performance metric; and
based on the performance of the two or more ensemble models, selecting the ensemble model for output.

### Claim 8 (depends on 1)

The system according to claim 1, wherein a record in the medical data is further indicative of one or more of: an identity of a surgeon, a clinical role of a surgeon, a type of surgical procedure, a surgical urgency, a type of post-surgery bed, an average surgery duration for a type of surgical procedure for a surgeon, and a number of surgical procedures performed during the surgery.

### Claim 9 (depends on 1)

The system according to claim 1, wherein the medical data further comprises patient data of patients of the respective surgeries.

### Claim 10 (depends on 9)

The system according to claim 9, wherein the patient data is indicative of one or more of: an age, a gender, a body-mass index, an American Society of Anaesthesiology (ASA)-score, a number of medications taken, a number of comorbidities, and a creatine level, of a respective patient.

### Claim 11 (depends on 1)

The system according to claim 1, wherein the processor subsystem is configured to identify the set of features in the training data using multivariate inferential analysis, wherein the processor subsystem is further configured to use univariate inferential analysis as a filter to determine an input set of features to the multivariate inferential analysis, wherein the features of the input set of features are individually predictive of the surgery duration.

### Claim 12 (depends on 1)

The system according to claim 1, wherein the predictive models comprise one or more of: a linear regression-based model, a random forest-based model, and a gradient boosting-based model.

### Claim 13 (depends on 1)

The system according to claim 1, wherein a record in the medical data is indicative of whether a surgery is an elective surgery or an acute surgery, wherein the processor subsystem is configured to generate different predictive models for predicting the surgical duration of elective surgeries and for predicting the surgical duration of acute surgeries.

### Claim 14

A computer-implemented method for generating a predictive model for predicting a surgery duration, comprising:
accessing medical data, the medical data comprising records of surgeries, wherein a record is indicative of at least a type of surgery and a surgery duration;
generating a predictive model for predicting the surgery duration by:
using a feature selection technique to identify a set of features in training data, which set of features is predictive of the surgery duration, wherein the training data comprises a first part of the medical data;
training a number of predictive models using the set of features in the training data as input and the surgery duration as prediction target, wherein the predictive models include at least a linear predictive model and a non-linear predictive model;
using a second part of the medical data, evaluating a performance of each the predictive models in predicting the surgery duration, wherein the evaluating of the performance comprises using a performance metric which characterises a time difference between a predicted surgery duration and an actual surgery duration;
based on the performance of the predictive models, generating an ensemble model which combines at least two of the predictive models; and
outputting the ensemble model for use in predicting the surgery duration.

### Claim 15 (depends on 14)

A transitory or non-transitory computer-readable medium comprising data representing a computer program, the computer program comprising instructions for causing a processor system to perform the method according to claim 14.
