This study will test how well an artificial intelligence (AI) software called Aorta AIM measures the size of the aorta (the body's main blood vessel) from CT scans. The aorta can become enlarged over time, which may lead to serious health problems. Doctors need to measure the aorta accurately to monitor patients and decide on treatment.
Currently, doctors measure the aorta manually on CT scans, which takes time and can vary between different doctors. Aorta AIM is designed to measure the aorta automatically and consistently.
In this study, researchers will compare Aorta AIM's measurements to those made by experienced radiologists or cardiologists (doctors who specialize in reading medical images). The study will use CT scans that have already been taken as part of routine medical care - no additional scans or procedures are needed.
The study will include approximately 250 participants across multiple hospitals in Brazil and the United States. Participants will be adults who have had a chest or abdominal CT scan that shows their aorta.
The main goals are to:
* Check if Aorta AIM measures the aorta as accurately as radiologists * See if Aorta AIM can help doctors work more efficiently * Evaluate if the software works well in different hospital settings and with different types of patients
This research may help improve how doctors monitor aortic disease and make treatment decisions in the future. There are no risks to participants since the study only uses existing medical images.
Adults ≥22 years at time of imaging
Studies performed 1 Jan 2016 - 31 August 2025
CT imaging of the whole, thoracic, or abdominal aorta, with or without contrast enhancement, acquired with any gating protocol (including non-gated)
Slice thickness ≤ 3mm
None
Adult patients (≥22 years of age) who underwent CT imaging of the aorta (whole, thoracic, or abdominal) as part of routine clinical care between January 1, 2016 and August 31, 2025. CT scans must include imaging of the aorta with or without contrast enhancement and have slice thickness ≤3mm. Patients with metal implants within 5cm of the aorta, prior aortic surgery, severe motion artifacts, congenital aortic anomalies, or post-traumatic aortic repairs are excluded. Patients who opted out of having their data used for medical research are also excluded. Retrospectively collected, de-identified CT imaging data from approximately 250 cases across multiple international sites in the United States and Latin America will be analyzed.
Case-only
Retrospective
Aorta AIM Automated Measurement
Diagnostic testAorta AIM is an investigational artificial intelligence software that automatically measures aortic diameters from CT imaging. The software analyzes retrospectively collected CT scans and generates standardized aortic diameter measurements at anatomically defined locations. These automated measurements will be compared to manual measurements performed by expert radiologists.
Expert Radiologist Manual Measurement
Diagnostic testManual aortic diameter measurements performed independently by two board-certified or equivalently experienced radiologists (minimum 5 years of cardiovascular imaging experience), each blinded to the other's measurements. Where the difference between the two reads does not exceed 3mm, the mean serves as the ground truth. Where the difference exceeds 3mm, a third senior reader performs independent adjudication, blinded to both primary reads. These consensus measurements serve as the reference standard for validation
Mean Absolute Error (MAE) between Aorta AIM measurements and expert reference standard
Mean Absolute Error (MAE) between aortic diameter measurements generated by Aorta AIM version 1.0 and the reference standard established by board-certified or equivalently experienced radiologists and/or cardiologists. Reference standard established through dual independent blinded reads with adjudication by a third senior reader where required. MAE will be calculated across all anatomical regions of the aorta assessed in the study. Acceptance criterion: MAE ≤ 5.0mm
Bland-Altman Limits of Agreement
Bland-Altman analysis to determine the limits of agreement between Aorta AIM automated measurements and expert manual measurements, providing assessment of systematic bias and random variation between the two measurement methods established through dual independent blinded reads with adjudication
Intraclass Correlation Coefficient (ICC)
Intraclass Correlation Coefficient assessing the reliability and agreement between Aorta AIM automated measurements and expert manual measurements across all anatomical regions of the aorta established through dual independent blinded reads with adjudication
Yield
Proportion of cases with valid ground truth for which Aorta AIM produces evaluable measurements at all predefined anatomical landmarks
Processing time
Time required for Aorta AIM software to process CT scans and generate automated aortic diameter measurements, measured in seconds or minutes from image input to final measurement output.
Click on the lead sponsor to view all of their trials.
