(Withdrawal) AI-Based Low-Dose 3D-DSA Reconstruction

(Withdrawal) Validation of a Patient-Specific Generative AI-Based Low-Dose Cerebrovascular 3D-DSA Image Reconstruction Method: a Stepwise, Multicenter, Randomized Crossover Trial

If the participants agree to participate in this study, the participants will undergo two scans (classic 3D-DSA and PS-3D-DSA assisted scan) to compare the imaging effects of both. After the procedure, the investigators will record the radiation exposure and collect DSA images.

Study Overview

• Status: Not yet recruiting
• Conditions: Cerebrovascular Disease
• Intervention / Treatment: Radiation: PS-3D-DSA; Radiation: classic 3D-DSA

Detailed Description

Although several previous studies have used deep learning methods to reduce 3D-DSA radiation dose, no prospective clinical trial had yet validated the practical application of these models. Herein, the investigators introduce a patient-specific generative AI-based low-dose cerebrovascular 3D-DSA image reconstruction method (PS-3D-DSA) to reconstruct 3D-DSA images from ultra-sparse 2D projection views and a prospective cohort is used to validate its efficacy in clinical practice.

• Study Type: Interventional
• Enrollment (Estimated): 134
• Phase: Not Applicable

Contacts and Locations

• Study Contact: Name: Huangxuan Zhao, PhD; Phone Number: +86 18627162379; Email: zhao_huangxuan@sina.com

Study Locations

• China
  • Hubei
    • Wuhan, Hubei, China, 430022
      • Wuhan Union Hospital

Participation Criteria

Eligibility Criteria

• Ages Eligible for Study: Adult; Older Adult
• Accepts Healthy Volunteers: Yes

Description

Inclusion Criteria:

1. Age ≥18 years.
2. Requires 3D-DSA-guided interventional diagnosis or treatment (e.g., cerebral angiography, cerebral artery chemoembolization) and meets operational indications.
3. Can understand the study's purpose, procedures, potential risks, and benefits, and voluntarily signs a written informed consent form.

Exclusion Criteria:

1. Severe heart or lung disease, such as heart failure or chronic obstructive pulmonary disease (COPD).
2. History of high-dose radiation exams or treatments.
3. Known allergies or severe adverse reactions to iodine contrast agents or other relevant medications.
4. Pregnant or breastfeeding women.
5. Severe comorbidities or chronic diseases (e.g., severe diabetes, renal insufficiency).
6. Severe mental illness or cognitive impairment preventing understanding of the study procedures or providing informed consent.

Study Plan

How is the study designed?

Design Details

• Primary Purpose: Diagnostic
• Allocation: Non-Randomized
• Interventional Model: Crossover Assignment
• Masking: Triple

Number of Arms

2

Arms and Interventions

Participant Group / Arm	Intervention / Treatment
Experimental: PS-3D-DSA; undergo a PS-3D-DSA scan	Radiation: PS-3D-DSA; undergo a PS-3D-DSA scan
Sham Comparator: classic 3D-DSA; undergo a classic 3D-DSA scan	Radiation: classic 3D-DSA; undergo a classic 3D-DSA scan

What is the study measuring?

Primary Outcome Measures

Outcome Measure	Measure Description	Time Frame
The radiation dose received by patients during interventional procedures when using two scanning protocols (classic 3D-DSA and PS-3D-DSA)	Using the built-in radiation monitoring function of interventional surgical equipment (DSA system), the radiation dose (AK，air kerma) received by the patient during the procedure is directly measured and recorded. The collected radiation dose data is documented in the patient's medical records and stored in the research database for subsequent analysis and comparative studies.	No more than 6 hours

Secondary Outcome Measures

Outcome Measure	Measure Description	Time Frame
The image diagnostic capabilities using two scanning protocols (classic 3D-DSA and PS-3D-DSA)	The secondary outcomes focus on evaluating the performance of interventional radiologists using either PS-3D-DSA or classic 3D-DSA for diagnostic tasks, with accuracy as the primary measure. Specifically, the diagnostic results from multiple radiologists will be compared against the gold standard to determine the level of agreement and diagnostic accuracy. This comparison will involve calculating sensitivity, specificity, and overall accuracy. Furthermore, receiver operating characteristic (ROC) curves will be plotted to assess the diagnostic performance and to visually represent the trade-off between sensitivity and specificity for each method.	No more than 1 month

Other Outcome Measures

Outcome Measure	Measure Description	Time Frame
Comparison of image quality using two scanning protocols (classic 3D-DSA and PS-3D-DSA)	Algorithmic performance for image quality and interventional radiologists' scoring of image. Specifically, multiple radiologists will provide subjective ratings of image quality, with the best quality rated as 5 and the worst as 1. Image Quality (5 points): point: The smoothness of the video is very poor, with noticeable stuttering in the animation, making it nearly impossible to provide a smooth viewing experience.; points: The smoothness of the video is poor, with a delayed animation, resulting in a less-than-ideal viewing experience.; points: The smoothness of the video is average, the animation is relatively smooth, offering a satisfactory viewing experience.; points: The smoothness of the video is high, with smooth and natural animation, providing an excellent viewing experience.; points: The smoothness of the video is extremely high, with exceptionally smooth animation, providing an outstanding viewing experience.	No more than 1 month

Collaborators and Investigators

• Sponsor: Wuhan Union Hospital, China

Publications and helpful links

General Publications

• van Asch CJ, Velthuis BK, Rinkel GJ, Algra A, de Kort GA, Witkamp TD, de Ridder JC, van Nieuwenhuizen KM, de Leeuw FE, Schonewille WJ, de Kort PL, Dippel DW, Raaymakers TW, Hofmeijer J, Wermer MJ, Kerkhoff H, Jellema K, Bronner IM, Remmers MJ, Bienfait HP, Witjes RJ, Greving JP, Klijn CJ; DIAGRAM Investigators. Diagnostic yield and accuracy of CT angiography, MR angiography, and digital subtraction angiography for detection of macrovascular causes of intracerebral haemorrhage: prospective, multicentre cohort study. BMJ. 2015 Nov 9;351:h5762. doi: 10.1136/bmj.h5762.
• Irfan M, Malik KM, Ahmad J, Malik G. StrokeNet: An automated approach for segmentation and rupture risk prediction of intracranial aneurysm. Comput Med Imaging Graph. 2023 Sep;108:102271. doi: 10.1016/j.compmedimag.2023.102271. Epub 2023 Jul 22.

Study record dates

Study Major Dates

• Study Start (Estimated): April 15, 2025
• Primary Completion (Estimated): May 31, 2025
• Study Completion (Estimated): May 31, 2025

Study Registration Dates

• First Submitted: January 2, 2025
• First Submitted That Met QC Criteria: January 7, 2025
• First Posted (Actual): March 25, 2025

Study Record Updates

• Last Update Posted (Actual): March 25, 2025
• Last Update Submitted That Met QC Criteria: March 13, 2025
• Last Verified: March 1, 2025

More Information

Terms related to this study

• Keywords: AI; Cerebrovascular; 3D-DSA
• Additional Relevant MeSH Terms: Brain Diseases; Central Nervous System Diseases; Nervous System Diseases; Vascular Diseases; Cardiovascular Diseases; Cerebrovascular Disorders
• Other Study ID Numbers: Patient-Specific Generative AI

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
• product manufactured in and exported from the U.S.: No