Quantification of Epicardial Adipose Tissue on CT
| Purpose | Automatically quantify and segment a patient’s epicardial adipose tissue on CT. |
| Tag(s) |
|
| Panel | Cardiac |
| Define-AI ID | 19040036 |
| Originator | Caterina Monti |
| Lead | Caterina Monti |
| Panel Chair | Carlo De Cecco |
| Panel Reviewers | Cardiac Panel |
| License | Creative Commons 4.0 |
| Status | Public Comment |
| RadElement Set | RDES85 |
Clinical Implementation
Value Proposition
Recently, research has reinforced the idea that epicardial adipose tissue (EAT) volume and density are associated with coronary artery disease (CAD), atrial fibrillation, and other adverse cardiovascular events. However segmenting the EAT on CT and measuring its volume and attenuation is a tedious and time-consuming task especially given the average slice thickness. This would be an excellent opportunity for an AI algorithm to assist clinicians and track EAT measurements that are otherwise left unreported. These data could help clinicians more appropriately identify and follow up patients at higher risk for adverse cardiovascular events from increased EAT. EAT could be assessed on noncontrast scans obtained for calcium scoring, thus effectively providing an additional biomarker for more accurate risk assessment.
Narrative(s)
Patient with intermediate cardiovascular risk according to the Framingham risk score undergoes CT comprising noncontrast images for evaluation of calcium scoring. Algorithm analyzes images and segments the EAT. With these measurements, potentially combined with calcium scoring, the clinician is able to establish a tailored follow-up scheme, based on individual cardiovascular risk.
Workflow Description
The image is obtained from modality and sent to PACS and the AI engine. The image is analyzed by the engine. The system then detects and estimates EAT volume and mean attenuation values. An alert message is sent to PACS from the engine with the information, identification, and graphic highlighting the EAT.
Considerations for Dataset Development
| EAT Location | surrounding the heart, more prominent at base and the apex, in the atrioventricular sulci, on the entire surface of the right ventricle, and around the great coronary vessels with their origins |
| EAT size | varied |
| Age | [18,100] |
| Sex at birth | Male, female |
| BMI | Varied |
| Ethnicity | varied |
| Arterial hypertension | Absent, present |
| Active smoking | Yes, No |
| Diabetes | Absent, present |
| Hypercholesterolemia | Absent, present |
| Cardiac height | varied |
| Arrhythmias | Absent, present |
| Framingham risk score | Any |
| Slice thickness | possibly lower than 1 mm |
Technical Specifications
Inputs
DICOM Study
| Procedure | CT (Calcium Scoring scans) |
| Views | Axial, coronal |
| Data Type | DICOM |
| Modality | CT |
| Body Region | Chest |
| Anatomic Focus | Heart |
Primary Outputs
EAT Volume
| RadElement ID | RDE483 |
| Definition | Volume of the epicardial adipose tissue |
| Data Type | Numeric |
| Value Set | N/A |
| Units | cm3 |
EAT Average Attenuation
| RadElement ID | RDE484 |
| Definition | Average attenuation of the epicardial adipose tissue |
| Data Type | Numeric |
| Value Set | N/A |
| Units | HU |
EAT Segmentation
| RadElement ID | RDE485 |
| Definition | Segmentation of the epicardial adipose tissue |
| Data Type | Coordinate |
| Value Set | N/A |
| Units | N/A |
Future Development Ideas
- The algorithm might compare patients’ EAT data over time, and provide an estimate of the variations in EAT.
- An algorithm for magnetic resonance segmentation might also be implemented, so that values can be compared between different techniques when patients undergo different examinations according to clinical indications.
- Additional measures such as EAT thickness estimated in a position equivalent to that commonly used for ultrasound could be implemented.