Machine Learning-assisted Analysis of Microcirculation Patterns and Parameters

Machine learning has been widely used in clinical medicine in recent years. It can be used for disease classification, disease severity grading, genetic testing, image analysis, adjuvant treatment recommendations, and predicting patient prognosis. Because sublingual microcirculation can be used for guiding shock resuscitation, a real time automated analysis is required for rapid changes of clinical condition. This study aims to use machine learning to analyze the parameters and patterns of sublingual microcirculation.

Study Overview

• Status: Recruiting
• Conditions: Sublingual Microcirculation Pattern and Parameter Analysis
• Intervention / Treatment: Device: sublingual microcirculation video recording

Detailed Description

The sublingual microcirculation videos are extracted from the 11 clinical trials conducting in the National Taiwan University Hospital.

In the first stage, the microcirculation videos and the related information are included in a de-identified manner. Each microcirculation video in the database will have a unique code. The video-related data will include the patient's height, weight, blood pressure, heartbeats, health status, major diseases, laboratory examination values, video quality description, automated vascular analysis (AVA) 3 software analysis results including total vessel density (TVD), perfused vessel density (PVD), proportion of perfused vessels (PPV), microvascular flow index (MFI), and heterogeneity index (HI). The length of each micro-cycle video is 4-6 seconds, and there are 25 frames per second. Take a picture as a representative image, each video can correspond to 4 images, and each micro-circulation image will also be marked with its image quality. Machine learning model will be trained for distinguishing the quality of videos and images. Only good-quality videos and images will be used for further analysis.

In the second stage, 80% of the microcirculation videos and images will be used for training and validation to find the best model, and then the remaining 20% of microcirculation videos and images will be used to test the model performance. The first training purpose is to automatically distinguish the size of blood vessels, calculate TVD, and draw a histogram of the number of microvessels of different diameters. The second training purpose is to measure the blood flow velocity in each small vessel and calculate PVD, MFI, and HI values.

• Study Type: Observational
• Enrollment (Anticipated): 800

Contacts and Locations

Study Locations

• Taiwan
  • Taipei, Taiwan, 10002
    • Recruiting
    • National Taiwan University Hospital

Participation Criteria

Eligibility Criteria

• Ages Eligible for Study: 20 years and older (ADULT, OLDER_ADULT)
• Accepts Healthy Volunteers: Yes
• Genders Eligible for Study: All

• Sampling Method: Non-Probability Sample

Study Population

Healthy volunteers Critically ill patients Surgical patients Patients with dialysis

Description

Inclusion Criteria:

• Microcirculation videos and images from previous clinical trials in the National Taiwan University Hospital with signed informed consent and agreement of further analysis

Exclusion Criteria:

• Microcirculation videos and images from previous clinical trials in the National Taiwan University Hospital with signed informed consent but disagreement of further analysis.

Study Plan

How is the study designed?

What is the study measuring?

Primary Outcome Measures

Outcome Measure	Measure Description	Time Frame
Perfused vessel density	Training machine learning models to view the videos of patients' sublingual microcirculation images and calculate the perfused vessel density. The videos of patients' sublingual microcirculation images are obtained and recorded by the video microscopes.	6 seconds

Secondary Outcome Measures

Outcome Measure	Measure Description	Time Frame
Patterns of microcirculation	Training machine learning models to view the videos of patients' microcirculation images and distinguish the patterns of microcirculation images and videos among healthy volunteers and patients with specific diseases or clinical conditions (eg. dialysis, postoperative, or septic shock.) The videos of patients' sublingual microcirculation images are obtained and recorded by the video microscopes.	6 seconds

Collaborators and Investigators

• Sponsor: National Taiwan University Hospital

Study record dates

Study Major Dates

• Study Start (ACTUAL): August 3, 2020
• Primary Completion (ANTICIPATED): December 1, 2023
• Study Completion (ANTICIPATED): April 1, 2024

Study Registration Dates

• First Submitted: June 23, 2021
• First Submitted That Met QC Criteria: June 30, 2021
• First Posted (ACTUAL): July 12, 2021

Study Record Updates

• Last Update Posted (ACTUAL): July 13, 2022
• Last Update Submitted That Met QC Criteria: July 11, 2022
• Last Verified: July 1, 2022

More Information

Terms related to this study

• Other Study ID Numbers: 202003094RINA

Plan for Individual participant data (IPD)

• Plan to Share Individual Participant Data (IPD)?: NO

Drug and device information, study documents

• Studies a U.S. FDA-regulated drug product: No
• Studies a U.S. FDA-regulated device product: No