Coronary Artery Disease
Conditions
Brief summary
Cardiac CT allows the assessment of the heart and of the coronary arteries by use of ionising radiation. Although radiation exposure was significantly reduced in recent years, further decrease in radiation exposure is limited by increased image noise and deterioration in image quality. Recent evidence suggests that further technological refinements with artificial intelligence allows improved post-processing of images with reduction of image noise. The present study aims at assessing the potential of a deep-learning image reconstruction algorithm in a clinical setting. Specifically, after a standard clinical scan, patients are scanned with lower radiation exposure and reconstructed with the DLIR algorithm. This interventional scan is then compared to the standard clinical scan.
Interventions
DEVICETrueFidelity
TrueFidelity (Deep Learning Image Reconstruction, DLIR) software by GE Healthcare. The medical device in question is a novel reconstruction algorithm for raw CT data which is based on artificial intelligence approaches, namely deep-learning iterative reconstruction (DLIR). This DLIR algorithm will be installed on the console of the CT Revolution scanning device, which is in routine clinical use for cardiac CT scans at the Department of Nuclear Medicine at the University Hospital Zurich. Purpose of this installation is the assessment of the performance of the DLIR algorithm during a limited time span of six weeks. The algorithm will be CE-marked at the time of installation and use (statement by GE Healthcare provided separately). Its intended use is the reconstruction of CT datasets. Of note, the novel DLIR algorithm will not substitute any clinical routine procedures currently in use. That is, diagnosis will still be made using the standard reconstruction algorithms.
Sponsors
University of Zurich
Study design
Allocation
NA
Intervention model
SINGLE_GROUP
Primary purpose
DIAGNOSTIC
Masking
NONE
Eligibility
Sex/Gender
ALL
Age
18 Years to No maximum
Healthy volunteers
No
Inclusion criteria
* Patients referred for cardiac CT angiography * Age ≥ 18 years * Written informed consent
Exclusion criteria
* Pregnancy or breast-feeding * Enrollment of the investigator, his/her family members, employees and other dependent persons * Renal insufficiency (GFR below 35 mL/min/1.73 m²)
Design outcomes
Primary
| Measure | Time frame | Description |
|---|---|---|
| Subjective Image Quality | Day 1 | Subjective image quality as measured by Likert scale from 1 (non-evaluable) to 5 (excellent) |
Secondary
| Measure | Time frame | Description |
|---|---|---|
| Signal Intensity | Day 1 | Signal intensity as average hounsfield units within a region of interest in the aortic root, change from experimental interventional to the control intervention |
| Image Noise | Day 1 | Image noise as standard deviation of hounsfield units within a region of interest in the aortic root, change from experimental interventional to the control intervention |
| Signal-to-noise Ratio | Day 1 | Signal-to-noise ratio |
| Dose-length Products | Day 1 | Comparison of dose-length products |
| Plaque Volumes | Day 1 | Quantitative analysis of coronary artery plaque volumes |
Countries
Switzerland
Participant flow
Recruitment details
Recruitment period: 08/May/2019 until 20/June/2019 Location: University Hospital Zurich, Department of Nuclear Medicine, Switzerland
Participants by arm
| Arm | Count |
|---|---|
| Normal-dose Versus Low-dose The standard intervention consists of the routinely performed cardiac CT datasets reconstructed with a standard iterative reconstruction algorithm (ASIR-V). Median radiation dose is about 0.5 mSv, range between about 0.2 and 1.2 mSv; median contrast agent administration about 45 mL, range between 35 and 55 mL.
The experimental intervention is an additional CT scan with a lower dose (about 20 to 50% decrease) and a similar contrast agent administration that is reconstructed with a deep-learning image reconstruction immediately after the clinical CT scan. The additional time required is about 5 minutes. | 50 |
| Total | 50 |
Baseline characteristics
| Characteristic | Normal-dose Versus Low-dose |
|---|---|
| Age, Continuous | 59 years STANDARD_DEVIATION 1 |
| Race (NIH/OMB) American Indian or Alaska Native | 0 Participants |
| Race (NIH/OMB) Asian | 0 Participants |
| Race (NIH/OMB) Black or African American | 0 Participants |
| Race (NIH/OMB) More than one race | 0 Participants |
| Race (NIH/OMB) Native Hawaiian or Other Pacific Islander | 0 Participants |
| Race (NIH/OMB) Unknown or Not Reported | 0 Participants |
| Race (NIH/OMB) White | 50 Participants |
| Region of Enrollment Switzerland | 50 participants |
| Sex: Female, Male Female | 7 Participants |
| Sex: Female, Male Male | 43 Participants |
Adverse events
| Event type | EG000 affected / at risk | EG001 affected / at risk |
|---|---|---|
| deaths Total, all-cause mortality | 0 / 50 | 0 / 50 |
| other Total, other adverse events | 0 / 50 | 0 / 50 |
| serious Total, serious adverse events | 0 / 50 | 0 / 50 |
Outcome results
Primary
Subjective Image Quality
Subjective image quality as measured by Likert scale from 1 (non-evaluable) to 5 (excellent)
Time frame: Day 1
| Arm | Measure | Group | Value (MEAN) | Dispersion |
|---|---|---|---|---|
| Normal-dose Versus Low-dose | Subjective Image Quality | Low-Dose | 5 Score on a Likert scale (0-5) 5=best | Standard Deviation 0 |
| Normal-dose Versus Low-dose | Subjective Image Quality | Normal-Dose | 5 Score on a Likert scale (0-5) 5=best | Standard Deviation 0 |
Secondary
Dose-length Products
Comparison of dose-length products
Time frame: Day 1
| Arm | Measure | Group | Value (MEDIAN) |
|---|---|---|---|
| Normal-dose Versus Low-dose | Dose-length Products | Low-Dose | 31 mGy*cm |
| Normal-dose Versus Low-dose | Dose-length Products | Normal-Dose | 52 mGy*cm |
p-value: <0.001Wilcoxon (Mann-Whitney)
Secondary
Image Noise
Image noise as standard deviation of hounsfield units within a region of interest in the aortic root, change from experimental interventional to the control intervention
Time frame: Day 1
| Arm | Measure | Group | Value (MEAN) | Dispersion |
|---|---|---|---|---|
| Normal-dose Versus Low-dose | Image Noise | Low-Dose | 27 Hounsfield units | Standard Deviation 4 |
| Normal-dose Versus Low-dose | Image Noise | Normal-Dose | 28 Hounsfield units | Standard Deviation 6 |
p-value: 0.9ANOVA
Secondary
Plaque Volumes
Quantitative analysis of coronary artery plaque volumes
Time frame: Day 1
| Arm | Measure | Group | Value (MEAN) | Dispersion |
|---|---|---|---|---|
| Normal-dose Versus Low-dose | Plaque Volumes | Low-Dose | 12.42 mm^3 | Standard Deviation 13.14 |
| Normal-dose Versus Low-dose | Plaque Volumes | Normal-Dose | 13.84 mm^3 | Standard Deviation 14.41 |
Secondary
Signal Intensity
Signal intensity as average hounsfield units within a region of interest in the aortic root, change from experimental interventional to the control intervention
Time frame: Day 1
| Arm | Measure | Group | Value (MEAN) | Dispersion |
|---|---|---|---|---|
| Normal-dose Versus Low-dose | Signal Intensity | Low-Dose | 462 Hounsfield units | Standard Deviation 76 |
| Normal-dose Versus Low-dose | Signal Intensity | Normal Dose | 443 Hounsfield units | Standard Deviation 85 |
p-value: 0.198ANOVA
Secondary
Signal-to-noise Ratio
Signal-to-noise ratio
Time frame: Day 1
| Arm | Measure | Group | Value (MEAN) | Dispersion |
|---|---|---|---|---|
| Normal-dose Versus Low-dose | Signal-to-noise Ratio | Low-Dose | 17 Ratio | Standard Deviation 3 |
| Normal-dose Versus Low-dose | Signal-to-noise Ratio | Normal-Dose | 16 Ratio | Standard Deviation 2 |
p-value: 0.8ANOVA