Research on AIS Recurrence Risk Prediction Model Using XGBoost Combined With Convolutional Neural Network Algorithm

The purpose of this observational study is to construct a recurrence risk prediction model for ischemic stroke within 1, 3, 6, and 12 months using XGBoost combined with Convolutional Neural Network (CNN) algorithm.

Method: Follow up was conducted on the study subjects at 1, 3, 6, and 12 months after discharge.

Follow up primary outcome: Whether the study subjects experienced recurrent stroke events.

Secondary outcome: Improved Rinkin score.

Collect information on research subjects:

It includes demographic data, physical examination, medical history, imaging images, medication use, scale scores, CYP2C19 genotype test results, laboratory tests, and other complex multidimensional data.

Study Overview

• Status: Completed
• Conditions: Ischemic Stroke

Detailed Description

This project conducts a one-year follow-up on the enrolled subjects (patients with ischemic stroke) to observe the recurrence and modified Rankin scores after discharge. Combining complex multidimensional data such as demographic information, physical examination, medical history, imaging images, medication status, NIHSS scale scores, Glasgow Coma Scale scores, CYP2C19 genotype test results, and laboratory examinations of the subjects; Convolutional Neural Networks (CNNs) are used to segment lesions and extract features from the subjects' imaging images; Cox regression models are employed to obtain factors influencing recurrence; a prediction model for the risk of recurrence within 1, 3, 6, and 12 months for ischemic stroke is constructed using the XGBoost algorithm combined with Convolutional Neural Networks. The exploration aims to provide new insights and methods for the prevention and control of major chronic diseases by evaluating the effectiveness of XGBoost in predicting the risk of ischemic stroke recurrence at different times.

1. Inclusion Criteria：

   • 18-85 years old;
   • Diagnosed with ischemic stroke or transient ischemic attack (diagnosis meets the criteria established by the Cerebrovascular Disease Group of the Neurology Branch of the Chinese Medical Association in 2014);
   • During the acute phase of onset (2 weeks);
   • Voluntarily participate and sign an informed consent form.

2. Exclusion criteria：

   • Cancer patients;
   • Cardiogenic infarction, cerebral infarction of other causes, and cerebral infarction of unknown causes;
   • Patients with hemorrhagic stroke, mixed stroke, and tumor stroke;
   • Merge with severe heart, lung, and liver system diseases;
   • Researchers identify patients with poor compliance and inability to complete long-term follow-up;
   • Patients currently participating in any clinical trials related to investigational drugs or medical devices.

• Study Type: Observational
• Enrollment (Actual): 2628

Contacts and Locations

Study Locations

• China
  • Jiangxi
    • Nanchang, Jiangxi, China, 330008
      • The Second Affiliated Hospital of Nanchang University

Participation Criteria

Eligibility Criteria

• Ages Eligible for Study: Adult; Older Adult
• Accepts Healthy Volunteers: No

• Sampling Method: Non-Probability Sample

Study Population

Ischemic stroke patients treated in hospitals who meet the inclusion and exclusion criteria for the project

Description

Inclusion Criteria:

• Aged between 18-85 years
• Diagnosed with ischemic stroke or transient ischemic attack (diagnosis in accordance with the standards set by the Cerebrovascular Disease Group of the Neurology Branch of the Chinese Medical Association in 2014)
• Within the acute phase of the illness (within 2 weeks)
• Voluntarily participates and signs an informed consent form

Exclusion Criteria:

• Patients with cancer
• Patients with cardiogenic infarction, other causes of cerebral infarction, or cryptogenic cerebral infarction
• Patients with hemorrhagic stroke, mixed stroke, and tumor stroke
• Patients with severe heart, lung, or liver system diseases
• Patients judged by the researcher to have poor compliance and unable to complete long-term follow-up
• Patients currently participating in any clinical trials related to investigational drugs or medical devices

Study Plan

How is the study designed?

What is the study measuring?

Primary Outcome Measures

Outcome Measure	Measure Description	Time Frame
Recurrent ischemic stroke,	Follow up was conducted on the study subjects at 1, 3, 6, and 12 months after discharge, with the main outcome being whether the study subjects experienced stroke recurrence events	April 6th, 2021- March 12th, 2024

Secondary Outcome Measures

Outcome Measure	Measure Description	Time Frame
modified Rankin score	Follow up was conducted on the study subjects at 1, 3, 6, and 12 months after discharge, with the secondary outcome being the modified Rinkin score.Improved Rinkin Scale: Completely asymptomatic (0 points). Despite symptoms, there is no obvious functional impairment and the ability to complete all daily work and life tasks (1 point). Mild disability, unable to complete all pre illness activities, but able to take care of daily affairs without assistance (2 points). Moderate disability, requiring partial assistance, but able to walk independently (3 points). Moderate to severe disability, unable to walk independently, requiring assistance from others in daily life (4 points). Severe disability, bedridden, urinary incontinence, and complete dependence on others in daily life (5 points). The higher the score, the worse the participant's neurological function recovery.	April 6th, 2021- March 12th, 2024

Collaborators and Investigators

• Sponsor: Second Affiliated Hospital of Nanchang University
• Investigators: Principal Investigator: yingping Y Yi, Second Affiliated Hospital of Nanchang University

Publications and helpful links

General Publications

• Liu J, Li J, Wu Y, Luo H, Yu P, Cheng R, Wang X, Xian H, Wu B, Chen Y, Ke J, Yi Y. Deep learning-based segmentation of acute ischemic stroke MRI lesions and recurrence prediction within 1 year after discharge: A multicenter study. Neuroscience. 2025 Jan 26;565:222-231. doi: 10.1016/j.neuroscience.2024.12.002. Epub 2024 Dec 2.

Helpful Links

• Explore deep - learning - based infarct lesion segmentation in AIS patients' brain MRI, radiomics - based 1 - year recurrence prediction, and develop a combined model for accurate AIS recurrence prediction

Study record dates

Study Major Dates

• Study Start (Actual): April 26, 2021
• Primary Completion (Actual): March 4, 2023
• Study Completion (Actual): December 31, 2024

Study Registration Dates

• First Submitted: January 13, 2025
• First Submitted That Met QC Criteria: January 21, 2025
• First Posted (Actual): March 25, 2025

Study Record Updates

• Last Update Posted (Actual): March 25, 2025
• Last Update Submitted That Met QC Criteria: February 19, 2025
• Last Verified: February 1, 2025

More Information

Terms related to this study

• Additional Relevant MeSH Terms: Cerebrovascular Disorders; Brain Diseases; Central Nervous System Diseases; Nervous System Diseases; Vascular Diseases; Cardiovascular Diseases; Pathologic Processes; Disease Attributes; Stroke; Ischemic Stroke; Recurrence
• Other Study ID Numbers: 2021efyB03

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