Advanced intelligent Clear-IQ Engine in Clinical Practice
Dr. John Hoe, Dr. Colin Tan
Introduction by Canon Medical
Parkway Radiology is a leading radiology and imaging provider in Singapore located at Mount Elizabeth Hospital, Mount Elizabeth Novena Hospital, Gleneagles Hospital and Parkway East Hospital. For more than 30 years, they have been synonymous with exceptional radiology services, delivering solutions to numerous healthcare providers in the country.
As a division under Singapore’s largest private healthcare operator, IHH Healthcare Singapore, Parkway Radiology leverages their proven track record for achieving medical excellence. Decades of experience enable them to be well poised in supporting the needs of the medical community not only in the country, but also in the region.
Parkway Radiology have eight Canon scanners in their fleet including Aquilion ONE / GENESIS Edition and Aquilion Prime SP.
Dr. John Hoe and Dr. Colin Tan outline their experiences with Canon’s Advanced intelligent Clear-IQ Engine (AiCE) in this white paper originally presented at SGCR Wires conference in September 2022.
Parkway Radiology is a leading radiology and imaging provider in Singapore located at Mount Elizabeth Hospital, Mount Elizabeth Novena Hospital, Gleneagles Hospital and Parkway East Hospital. For more than 30 years, they have been synonymous with exceptional radiology services, delivering solutions to numerous healthcare providers in the country.
As a division under Singapore’s largest private healthcare operator, IHH Healthcare Singapore, Parkway Radiology leverages their proven track record for achieving medical excellence. Decades of experience enable them to be well poised in supporting the needs of the medical community not only in the country, but also in the region.
Parkway Radiology have eight Canon scanners in their fleet including Aquilion ONE / GENESIS Edition and Aquilion Prime SP.
Dr. John Hoe and Dr. Colin Tan outline their experiences with Canon’s Advanced intelligent Clear-IQ Engine (AiCE) in this white paper originally presented at SGCR Wires conference in September 2022.
Dose Awareness
CT scanning has revolutionized diagnosis and treatment, almost eliminating the need for exploratory surgeries and many other invasive and potentially risky procedures. The benefits of CT exams, when appropriate, far outweigh any radiation-associated cancer risks, and the risk from a single CT scan is quite small1. However, even though the risk is small, it is important to reduce radiation dose as much as possible for each examination1. The challenge is to reduce radiation dose without compromising on image quality or reducing diagnostic confidence.
In clinical practice, a dose reduction strategy includes working with referrers on alternative imaging techniques and protocols to minimize radiation dose, working with technologists to ensure they adhere to best practice protocols and, finally, the implementation of the latest dose reduction technology available on each scanner.
Canon Medical has developed the next generation in dose reduction technology utilizing Artificial Intelligence.
In clinical practice, a dose reduction strategy includes working with referrers on alternative imaging techniques and protocols to minimize radiation dose, working with technologists to ensure they adhere to best practice protocols and, finally, the implementation of the latest dose reduction technology available on each scanner.
Canon Medical has developed the next generation in dose reduction technology utilizing Artificial Intelligence.
Advanced intelligent Clear-IQ Engine – How it Works
AiCE is a Deep Learning Reconstruction algorithm (DLR) that takes advantage of the enormous computational power of a Deep Learning Neural Network to dramatically reduce noise while maintaining or improving diagnostic resolution in the resultant images. Therefore, this AI-based reconstruction technique allows CT examinations to be performed with significantly less radiation dose.
Deep Learning is a type of machine learning that uses multi-layered neural networks to perform a task. Neural networks are comprised of thousands of “neurons,” each of which perform a mathematical operation on the image data. The computational power behind thousands of neurons working together allows the network to generate more sophisticated rules for identifying signal features than is possible with conventionally programmed reconstruction. In the case of AiCE, the task the network learns to perform is simply to distinguish signal features from noise. It identifies and enhances signal while removing noise, resulting in an industry-first low contrast specification of 1.5 mm at 3 HU and 22.6 mGy2.
Deep Learning is a type of machine learning that uses multi-layered neural networks to perform a task. Neural networks are comprised of thousands of “neurons,” each of which perform a mathematical operation on the image data. The computational power behind thousands of neurons working together allows the network to generate more sophisticated rules for identifying signal features than is possible with conventionally programmed reconstruction. In the case of AiCE, the task the network learns to perform is simply to distinguish signal features from noise. It identifies and enhances signal while removing noise, resulting in an industry-first low contrast specification of 1.5 mm at 3 HU and 22.6 mGy2.
The key to a successful Deep Learning neural network lies in how it is trained, i.e. the process by which the neurons learn to perform the desired task. In order to learn, the network is provided with pairs of data, a low-quality image and a gold-standard reference image. The network adjusts the weights of its neurons until the difference between the input low-quality image and the gold standard image is minimized. In the case of AiCE, the gold-standard clinical reference images are acquired with high tube current and reconstructed with a Model-Based Iterative Reconstruction (MBIR), that takes into account modelling of system optics, system physics, scanner statistical properties and human anatomy, and uses a greater number of iterations than could be otherwise used in a clinical setting due to time constraints2.
Once the network is trained and validated, the AiCE reconstruction is implemented into the scanner, where it does not continue to learn (Figure 1). The AiCE reconstruction is included in the scan protocol for seamless workflow.
Once the network is trained and validated, the AiCE reconstruction is implemented into the scanner, where it does not continue to learn (Figure 1). The AiCE reconstruction is included in the scan protocol for seamless workflow.
Figure 1: AiCE algorithm training and implementation.
DLR Implementation
AiCE is available on all of our Canon CT systems, not just restricted to the high-end scanners within our network. As AiCE is integrated into the automatic exposure control (SUREExposure) it reduces the X-ray tube current automatically to reduce dose by as much as 82% compared to Filtered Back Projection (FBP). Our radiographers do not have to perform any additional post processing when using AiCE as it is implemented into all our scan protocols. Reconstruction speed is also important when implementing new reconstruction technology into our practice. On our Prime SP scanner, the reconstruction speed of AiCE is 33 sec compared to 20 sec with AIDR (Adaptive Iterative Dose Reduction) 3D for a soft tissue reconstruction with a range of 216mm. In routine clinical practice this difference is negligible and has no effect on our patient throughput.
DLR for Cardiac Imaging
Figure 2: Improved visualization of plaque and vessel lumen with AiCE compared to AIDR 3D.
At Parkway we have used AiCE since the end of 2019 and have performed an evaluation of our initial cases. 87 control cases using AIDR 3D and 306 study cases using AiCE were analyzed for signal-to-noise ratio (SNR), contrast noise ratio (CNR), image quality and radiation dose. The results are shown in Table 1. AiCE resulted in an increase in SNR of 35.3% and CNR of 40.5%. The radiation dose decreased by 23% compared with cases scanned using AIDR 3D. All patients in this evaluation included patients with arrhythmia, high heart rates and the majority were scanned using a 40-80% exposure window. Figure 2 shows an example of the image quality improvements seen in this study. AiCE provides reduced noise, sharper vessel wall, improved visualization of the vessel lumen, plaque and stents.
Figure 2: Improved visualization of plaque and vessel lumen with AiCE compared to AIDR 3D.
| AIDR 3D (n=87) | AiCE (n=306) | % change | |
|---|---|---|---|
| CTDIvol (mGy) | 16.0 | 12.0 | -25 |
| DLP (mGy•cm) | 217.0 | 169.0 | -22 |
| Effective Dose (mSv)* | 3.0 | 2.3 | -23 |
| SNR | 14.7 | 19.9 | +35.3 |
| CNR | 13.1 | 18.4 | +40.5 |
ECG Volume scanning with AiCE can be used in combination with Single Energy Metal Artifact Reduction (SEMAR) to improve visualization of the heart in patients with pacemakers and defibrillator leads. In Figure 3 the right coronary artery is obscured in the image without SEMAR.
After reconstruction with SEMAR and AiCE the artery can be visualized.
After reconstruction with SEMAR and AiCE the artery can be visualized.
Figure 3: Example of a cardiac scan reconstructed with and without SEMAR in a patient with a pacemaker lead.
DLR for Body Imaging
Figure 4: Results of evaluation of liver exams comparing AIDR 3D and AiCE. Black lines = median.
We performed an evaluation of our initial abdomen cases where each case was reconstructed with AIDR 3D and AiCE at two different levels of noise reduction, AiCE mild and AiCE standard. 100 cases were analyzed for image noise (SD), signal-to-noise ratio (SNR) and contrast noise ratio (CNR). The results are shown in Figure 4 and Table 2. We found that AiCE standard resulted in a 32% reduction in noise, 51% increase in SNR and 48.7% increase in CNR when compared to AIDR 3D.
Figure 4: Results of evaluation of liver exams comparing AIDR 3D and AiCE. Black lines = median.
| AIDR 3D | AiCE Mild | AiCE Standard | |
|---|---|---|---|
| Noise | 15.2 | 12.7 (16.4%) | 10.25 (32%) |
| SNR | 6.95 | 8.46 (22.3%) | 10.53 (51%) |
| CNR | 3.24 | 3.88 (19.5%) | 4.82 (48.7%) |
Clinical Examples
The following examples demonstrate AiCE in abdominal examinations.
Case 1: This 64 kg man presented for an abdominal scan for possible bowel obstruction. Liver hypodensities are clearly seen in the venous phase (red arrow). Gallbladder wall adenomyomatosis is also seen. The renal cyst (blue arrow) was determined to be a Bosniak1 type cyst.
CTDI: 3.3 mGy, DLP: 172.2 mGy•cm, Effective Dose 2.6 mSv (k=0.015). Scanner: Aquilion Prime SP.
Case 2: This 75 kg woman underwent an abdomen examination. In the pelvis there is a complex cystic lesion confirmed on MRI. AiCE images provide clear visualization of pelvic structures.
CTDI: 4.8 mGy, DLP: 251.8 mGy•cm, Effective Dose 3.8 mSv (k=0.015). Scanner: Aquilion Prime SP.
Conclusion
AiCE Deep Learning Reconstruction can be routinely used in daily clinical practice for all body regions. The integration into the scan protocols provides patient-specific dose reduction without requiring any additional work for the radiographers. The reconstruction speed is fast enough to be used in daily workflow. At Parkway, AiCE has provided improvements in image quality and significant dose reduction in both cardiac and liver examinations and has been implemented into our routine clinical protocols to realize low radiation dose for each patient while maintaining image quality.
Using Canon’s AiCE reconstruction for cardiac and body imaging brings significant opportunities to reduce dose and improve image quality.//
Using Canon’s AiCE reconstruction for cardiac and body imaging brings significant opportunities to reduce dose and improve image quality.//
References
1. https://www.health.harvard.edu/cancer/radiation-risk-from-medical-imaging
2. Boedecker K, Advanced intelligent Clear-IQ Engine (AiCE) Deep Learning Reconstruction: Whole Organ Coverage with
extraordinary image quality at the fast pace of Medicine, White Paper, 2020
Dr. John Hoe
Consultant Radiologist Parkway Radiology, Singapore
Dr. Colin Tan
Consultant Nuclear Medicine Physician Parkway Radiology, Singapore
1. https://www.health.harvard.edu/cancer/radiation-risk-from-medical-imaging
2. Boedecker K, Advanced intelligent Clear-IQ Engine (AiCE) Deep Learning Reconstruction: Whole Organ Coverage with
extraordinary image quality at the fast pace of Medicine, White Paper, 2020
Dr. John Hoe
Consultant Radiologist Parkway Radiology, Singapore
Dr. Colin Tan
Consultant Nuclear Medicine Physician Parkway Radiology, Singapore
