UMIN-CTR Clinical Trial

Public title

Effective lung protective ventilation strategy using deep learning with graphic monitors : Development of an artificial intelligence prediction model.

Acronym

ELPIS-grad STUDY

Scientific Title

Effective lung protective ventilation strategy using deep learning with graphic monitors : Development of an artificial intelligence prediction model.

Scientific Title:Acronym

ELPIS-grad STUDY

Region

Japan

Condition

Ventilated patients admitted to the Yamagata University Hospital Advanced Intensive Care Center

Classification by specialty

Anesthesiology	Emergency medicine	Intensive care medicine

Classification by malignancy

Others

Genomic information

NO

Narrative objectives1

Reading graphic monitors requires considerable clinical experience and a high level of expertise in respiratory physiology.
The main objective of this study is to develop an AI system that predicts the need for interventions such as changing ventilatory settings by creating AI models from graphic monitor images acquired from ventilators of patients undergoing ventilatory therapy in intensive care units.

Basic objectives2

Efficacy

Basic objectives -Others

Trial characteristics_1

Trial characteristics_2

Developmental phase

Primary outcomes

Compare the AI system's prediction of the need for setting changes and reasons for setting changes with the actual presence or absence of setting changes and reasons for setting changes by the intensivist. From that comparison, ROC curves are drawn and accuracy, sensitivity, specificity, and AUC are calculated.

Key secondary outcomes

Study type

Observational

Basic design

Randomization

Randomization unit

Blinding

Control

Stratification

Dynamic allocation

Institution consideration

Blocking

Concealment

No. of arms

Purpose of intervention

Type of intervention

Interventions/Control_1

Interventions/Control_2

Interventions/Control_3

Interventions/Control_4

Interventions/Control_5

Interventions/Control_6

Interventions/Control_7

Interventions/Control_8

Interventions/Control_9

Interventions/Control_10

Age-lower limit

20

years-old

<=

Age-upper limit

99

years-old

>=

Gender

Male and Female

Key inclusion criteria

Ventilated patients admitted to the Advanced Intensive Care Center, Yamagata University Hospital

Key exclusion criteria

None

Target sample size

200

Name of lead principal investigator

1st name	Masaki
Middle name
Last name	Nakane

Organization

Yamagata University Hospital

Division name

Department of emergency medicine

Zip code

990-2331

Address

2-2-2, Iida-Nishi, Yamagata City, Yamagata Prefecture

TEL

0236285400

Email

hayasakatatsuya1101@gmail.com

Name of contact person

1st name	Tatsuya
Middle name
Last name	Hayasaka

Organization

Yamagata University Hospital

Division name

Department of Anesthesiology

Zip code

990-2331

Address

2-2-2, Iida-Nishi, Yamagata City, Yamagata Prefecture

TEL

0236285400

Homepage URL

Email

hayasakatatsuya1101@gmail.com

Institute

Yamagata Universal Faculty of Medcine

Institute

Department

Personal name

Organization

Japan Society for the Promotion of Science

Organization

Division

Category of Funding Organization

Japanese Governmental office

Nationality of Funding Organization

Co-sponsor

Name of secondary funder(s)

Organization

The Ethical Review Committee of Yamagata University Faculty of Medicine

Address

2-2-2, Iida-Nishi, Yamagata City, Yamagata Prefecture

Tel

0236285015

Email

ikekenkyu@jm.kj.yamagata-u.ac.jp

Secondary IDs

NO

Study ID_1

Org. issuing International ID_1

Study ID_2

Org. issuing International ID_2

IND to MHLW

Institutions

Date of disclosure of the study information

2023

Year

08

Month

20

Day

URL releasing protocol

Publication of results

Unpublished

URL related to results and publications

Number of participants that the trial has enrolled

Results

Results date posted

Results Delayed

Results Delay Reason

Date of the first journal publication of results

Baseline Characteristics

Participant flow

Adverse events

Outcome measures

Plan to share IPD

IPD sharing Plan description

Recruitment status

Preinitiation

Date of protocol fixation

2023

Year

08

Month

18

Day

Date of IRB

Anticipated trial start date

2023

Year

08

Month

20

Day

Last follow-up date

2026

Year

03

Month

31

Day

Date of closure to data entry

Date trial data considered complete

Date analysis concluded

Other related information

1) To collect clinical data from eligible patients as a prospective study.
Methods: Ventilator graphic image data were obtained. When ventilator mode or settings are changed by an intensivist skilled in respiratory therapy, the setting changes and the reasons for the changes are recorded.

2) Analysis of the captured clinical data and the need for modification.
Method: Using captured graphic images as input values, machine learning is performed to link the presence or absence of ventilatory setting changes and the reasons for the setting changes. The presence or absence of the setting change and the reason for the setting change are used as output values to predict the necessity of the change.

3) System construction
1) Construct an AI prediction model using the collected graphic monitor images.
2) Construct an AI system for predicting changes in ventilatory settings using train data (80% of the total data), and deep learning, transfer learning, and fine tuning using train data.
(3) Verify the accuracy of the constructed system using 20% of the data as test data.
4) Verify the prediction by the AI system and the setting change by the intensivists, and calculate AUC based on accuracy, sensitivity, specificity, and ROC curve.
(5) Visualize the evaluation area of the model with a class activation heat map (Grad-CAM) of the Test data after the AI model is created in order to clarify which area the AI system focuses on for evaluation.

Registered date

2023

Year

08

Month

18

Day

Last modified on

2023

Year

08

Month

18

Day

Link to view the page

Value
https://center6.umin.ac.jp/cgi-open-bin/ctr_e/ctr_view.cgi?recptno=R000059275