Stepped-Wedge Cluster Randomized Trial of AI-Assisted CTA Detection for Intracranial Aneurysms in Regional Hospitals

NCT ID: NCT07124624

Last Updated: 2025-08-20

Basic Information

• Recruitment Status: NOT_YET_RECRUITING
• Clinical Phase: NA
• Total Enrollment: 14400 participants
• Study Classification: INTERVENTIONAL
• Study Start Date: 2025-10-09
• Study Completion Date: 2029-08-31

Brief Summary

This study (IDEAL 2) is a nationwide stepped-wedge cluster-randomized trial designed to prospectively enroll over 14,400 patients undergoing outpatient head CT angiography (CTA). The trial will be conducted across more than 72 regional hospitals in China. Clusters were randomly assigned to nine randomization groups. In accordance with the stepped-wedge design, clusters will sequentially transition from the control condition (standard human diagnosis) to the intervention condition (AI-assisted diagnosis) at regular intervals over a 10-month period, until all clusters receive the intervention. The primary outcome is the detection rate of intracranial aneurysms. Secondary outcomes include patient prognosis and clinical outcomes.

Detailed Description

A multicenter, stepped-wedge cluster-randomized trial will be conducted in regional hospitals, specifically prefecture-level and county-level institutions across China. Each cluster (i.e., hospital) will enroll approximately 200 patients undergoing head computed tomography angiography (CTA), yielding a total sample size of at least 14,400 participants. The trial consists of nine steps, each lasting one month. Clusters will transition sequentially from the control condition to the intervention condition based on stratified randomization, until all clusters have received the intervention.

In the control group, diagnoses and treatments will follow local standard clinical protocols. In the intervention group, diagnostic procedures will be supported by an artificial intelligence (AI)-assisted system. The primary outcome is the detection rate of intracranial aneurysms, as determined from radiology reports at the patient level. Secondary outcomes include additional diagnostic performance metrics on CTA, such as the detection of intracranial arterial stenosis, occlusion, and tumors.

Follow-up evaluations at 3 and 12 months will assess treatment-related indicators-including repeat head CTA or magnetic resonance angiography (MRA), hospitalization rates, and digital subtraction angiography (DSA) utilization-as well as clinical outcomes related to aneurysm events. These measures aim to evaluate both the short- and long-term impacts of AI-assisted diagnosis on routine clinical practice and patient prognosis.

Conditions

Intracranial Aneurysm; CT Angiography; AI (Artificial Intelligence); Cluster Randomized Trial

Study Design

• Allocation Method: RANDOMIZED
• Intervention Model: SEQUENTIAL
• Primary Study Purpose: DIAGNOSTIC
• Blinding Strategy: NONE

Study Groups

AI-Assisted Diagnosis

Patients in this group will undergo head CTA, interpreted by radiologists with the assistance of a previously validated deep learning AI model. The model highlights suspected aneurysm candidates on the CTA images to support diagnostic decision-making.

Group Type: EXPERIMENTAL

AI-Assisted CTA Interpretation

Intervention Type: DEVICE

A locked, independently validated deep learning model was used to assist radiologists in interpreting head CTA scans. The model was trained on 16,546 CTA cases and externally validated on an independent set of 900 DSA-verified CTA cases, achieving a patient-level sensitivity of 0.943 and an average of 0.187 false positives per case.

Standard Diagnosis

Patients in this group will undergo head CTA, which will be interpreted by radiologists following standard local diagnostic protocols. No AI assistance will be provided during image interpretation.

Group Type: ACTIVE_COMPARATOR

Standard CTA Interpretation

Intervention Type: DIAGNOSTIC_TEST

Head CTA interpretation performed by radiologists using local routine diagnostic workflows without AI support.

Interventions

AI-Assisted CTA Interpretation

A locked, independently validated deep learning model was used to assist radiologists in interpreting head CTA scans. The model was trained on 16,546 CTA cases and externally validated on an independent set of 900 DSA-verified CTA cases, achieving a patient-level sensitivity of 0.943 and an average of 0.187 false positives per case.

Intervention Type: DEVICE

Standard CTA Interpretation

Head CTA interpretation performed by radiologists using local routine diagnostic workflows without AI support.

Intervention Type: DIAGNOSTIC_TEST

Eligibility Criteria

Inclusion Criteria

-Patients in the outpatient setting who are scheduled to undergo head CTA scanning

Exclusion Criteria

• Age < 18 years
• History of cerebrovascular surgery involving any metallic implants (e.g., aneurysm embolization, aneurysm clipping, or vascular stenting)
• Modified Rankin Scale (mRS) score > 3
• Refuse to sign written informed consent
• Contraindications to CTA examination
• CTA scan failure, incomplete imaging data, or image quality insufficient for diagnostic evaluation

• Minimum Eligible Age: 18 Years
• Eligible Sex: ALL
• Accepts Healthy Volunteers: Yes

Sponsors

Jinling Hospital, China

OTHER

Sponsor Role: lead

Responsible Party

Zhang longjiang,MD

Head of Radiology

Responsibility Role: PRINCIPAL_INVESTIGATOR

Central Contacts

Longjiang Zhang, Ph.D, MD

Role: CONTACT

kevinzhlj@163.com

+8613405833176

Bin Hu, MS

Role: CONTACT

hubin011601@163.com

+8618851088705

Other Identifiers

2025DZKY-087-01

Identifier Type: -

Identifier Source: org_study_id