Coronary Artery Disease, Coronary Atheroscleroses
Conditions
Keywords
Artificial intelligence, Radiomics, Deep learning, Coronary artery disease, Non-contrast computed tomography
Brief summary
This project aims to improve direct patient care by reducing the risks of futile exposure to ionizing radiation and iodinated contrast in patients referred for coronary computed tomography angiography
Detailed description
Since the last NICE guidelines update recommending computed tomography coronary angiography (CTCA) as the first line of investigation for patients with suspected coronary artery disease (CAD), there has been a high burden in the healthcare system and unnecessary exposition to radiation and iodine-containing contrast medium, especially in the youngest. Around 35% of patients who currently undergo CTCA have normal coronaries which means those patients were unnecessary exposed to radiation and contrast. A CTCA screening strategy to rule out CAD is needed to comply with the ALARA (As Low As Reasonable Achievable) principles preventing radiation risks, reducing unnecessary scans and directing healthcare resources to those who will benefit from a CTCA. We designed the SAFE-CT (Screening coronary Artery disease using artiFicial intelligencE in noncontrast Computed Tomography) study to develop a state-of-art artificial intelligence method to detect CAD as defined on CTCA using high-dimensional data (radiomics) extracted from the non-contrast cardiac computed tomography (CT). The model will be trained in 15,000 subjects scanned with paired non-contrast CT and CTCA and externally validated in an independent cohort of 1,000 subjects. In a preliminary analysis, non-contrast CT radiomics improved calcium score performance and discriminated CAD with an AUC of 0.91 (95% CI: 0.83-1.00). The algorithm will be converted into a user-friendly plugin to automatically decide whether the patient needs contrast. A real-world multicentre cohort study will be planned for software prospective validation and the creation of a large-scale proteomic biobank to support the translation of imaging biomarkers worldwide. SAFE-CT can change the current CT scanning workflow by creating software that accurately rules out any CAD in \>1/3 of patients referred for CTCA with low radiation and no contrast. This accurate machine learning model will be optimized to reach \>90% sensitivity and negative predictive value and will bring several advantages for patients and the healthcare system: * Prevention of radiation and contrast exposition. * Increased CTCA scanning capacity for complex cases. * Widespread use of CT for CAD exclusion in the emergency department and in outpatient clinics of centres with no CTCA. * Improved screening tool for CAD in asymptomatic subjects. * Up- and downstream cost reduction. The SAFE-CT project proposes a safer, low-cost, and personalized CTCA scanning strategy that fosters scientific and technological innovation with the potential to bring improvement to patient care and clinical practice, and, thereby, societal, and economic impact.
Interventions
DIAGNOSTIC_TESTCT coronary angiography and non-contrast CT
A CTCA is an X-ray computed tomography of the coronary arteries that allows visualization of coronary plaques with high temporal and spatial resolution, however, it implies the use of iodine contrast and exposition to clinically significant ionizing radiation. Non-contrast ECG-gated CT (calcium score - CCS image). A non-contrast cardiac CT for CCS can be performed very quickly with significantly lower radiation (\ 6 times lower) than CTCA and without the need for contrast.
Sponsors
University of Oxford
University of Edinburgh
Universidade do Porto
Study design
Observational model
OTHER
Time perspective
RETROSPECTIVE
Eligibility
Sex/Gender
ALL
Age
18 Years to 100 Years
Healthy volunteers
Yes
Inclusion criteria
\- Patient with stable chest pain who underwent a CTCA
Exclusion criteria
* Missing non-contrast CT image (coronary calcium score image) * Known coronary artery disease * Prior myocardial infarction * Prior PCI or CABG
Design outcomes
Primary
| Measure | Time frame |
|---|---|
| Create a national registry of cardiac CT | 3 years |
| Build a user-friendly plugin to facilitate users experience and distribution of our technology in clinical practice. | 3 years |
| Evaluate the real-world operationality and performance of the plugin in an international multicentre prospective cohort study. | 3 years |
| Implement a machine learning model to detect coronary inflammation as defined using the Fat Attenuation Index (FAI ≥ -70.1 HU) in patients with no visible coronary plaque (CAD-RADS=0). | 3 years |
| Build a non-contrast CT radiomic signature of CAD | 3 years |
| Implement a machine learning model to discriminate patients with no CAD from patients with at least minimal disease (CAD-RADS=0 vs. CAD-RADS>0). | 3 years |
Secondary
| Measure | Time frame |
|---|---|
| Setup a public CT imaging repository | 3 years |
| Setup a human blood biobank to identify the peripheral blood mononuclear cells (PBMCs) and plasma proteomics associated with CT data and clinical outcomes. | 3 years |
Countries
Portugal
Outcome results
None listed