Development and Application of an AI Model for Accurate Interpretation of Abdominal Enhanced CT Images
Trial Snapshot
- Phase
- Not Applicable
- Status
- Active, not recruiting
- Sponsor
- Enrollment
- 2,000,000
- Locations
- 1
- Primary Endpoint
- Performance of AI Model for Lesion Detection on Abdominal Contrast-Enhanced CT
Study Overview
Brief Summary
This study aims to develop an AI-assisted diagnostic system for abdominal contrast-enhanced CT images using data from multiple inpatient centers. In collaboration with Alibaba DAMO Academy, the project will address key mathematical challenges limiting current automated image interpretation, including feature space alignment, hybrid reasoning, and multimodal report generation. The study includes the following components: (1) construction of a dual-modality foundation model to align abdominal CT features with corresponding radiology reports; (2) development of a model to standardize CT phase variation among patients; and (3) creation of an automated image interpretation and reporting system that integrates multi-source clinical data. The effectiveness of the system will be evaluated through a report quality assessment framework and clinical validation. This project aims to improve the accuracy and clinical applicability of automated abdominal disease interpretation and promote intelligent innovation in healthcare delivery.
Study Design
- Study Type
- Observational
- Observational Model
- Cohort
- Time Perspective
- Retrospective
Eligibility Criteria
- Ages
- 18 Years to — (Adult, Older Adult)
- Sex
- All
- Accepts Healthy Volunteers
- Yes
Inclusion Criteria
- •multiphase contrast-enhanced abdominal CT covering the full abdominal region and corresponding radiology reports matched to the CT images
Exclusion Criteria
- •CT images with poor diagnostic quality due to artifacts, including but not limited to: Convolution artifacts caused by improper arm positioning (e.g., arms placed alongside the body instead of above the head),Respiratory motion artifacts due to inadequate breath-holding.
Arms & Interventions
Internal Training Set
Internal Validation Set
External Test Set
Outcomes
Primary Outcomes
Performance of AI Model for Lesion Detection on Abdominal Contrast-Enhanced CT
Time Frame: After internal and external validation datasets are processed (estimated 6-12 months)
The primary outcome is the overall performance of the AI model in detecting and characterizing lesions in abdominal organs using multiphase contrast-enhanced CT scans. Performance will be measured using area under the receiver operating characteristic curve (AUC), F1-score, sensitivity, and specificity, with expert radiologist consensus reports as the reference standard.
Secondary Outcomes
No secondary outcomes reported
Investigators
Qi Zhang
Professor
First Affiliated Hospital of Zhejiang University