HYPER DLR
by Shanghai United Imaging Healthcare Co.
· Based in China →
— Leading Healthcare Innovation, Equal Healthcare for All
Included with system purchase and service contract
Regulatory Status Disclosed
Overview
HYPER DLR (Deep Learning Reconstruction) is an advanced artificial intelligence (AI) powered image processing function developed by Shanghai United Imaging Healthcare Co. It is specifically designed for use with Positron Emission Tomography (PET) and Computed Tomography (CT) systems. The tool utilizes a deep learning technique in post-processing to significantly reduce noise in fluorodeoxyglucose (FDG) PET images. By extracting noise components from the original image, HYPER DLR improves the signal-to-noise ratio (SNR) and contrast-to-noise ratio (CNR) while maintaining image consistency. This enhancement in image quality is crucial for accurate diagnosis, and it also enables the possibility of shorter scanning times and/or reduced radiation doses for patients during PET/CT exams.
Reviewed by Pouyan Golshani, MD — Interventional Radiologist
Key Features
- Deep Learning Reconstruction (DLR) technology
- Reduces noise in FDG PET images
- Improves Signal-to-Noise Ratio (SNR)
- Improves Contrast-to-Noise Ratio (CNR)
- Preserves image consistency
- Enables shorter PET/CT scan times
- Allows for lower radiation doses in PET/CT exams
- Software-only post-processing technique
- Integrated with United Imaging PET/CT systems (e.g., uMI 550, uMI 780, uEXPLORER, uMI Panvivo)
- Part of the 'Software Upgrades for Life' program for existing customers
Use Cases
- Enhancing diagnostic image quality in PET/CT scans for clearer visualization
- Reducing patient radiation exposure during PET/CT examinations
- Shortening PET/CT scan times to improve patient comfort and throughput
- Improving lesion detectability in PET images
- Assisting in the detection, localization, diagnosis, staging, and restaging of diseases
- Supporting treatment planning and evaluation of treatment response in oncology, cardiology, and neurology
What Physicians Need to Know
Key Capabilities
HYPER DLR is a Deep Learning Reconstruction (DLR) technology designed to reduce noise in Fluorodeoxyglucose (FDG) PET images. It significantly improves Signal-to-Noise Ratio (SNR) and Contrast-to-Noise Ratio (CNR) by up to 42%, while preserving image consistency. The technology utilizes a pre-trained neural network to extract and subtract noise components, and it is built on a UNet architecture with residual and dense connections.
Clinical Utility
This tool enhances diagnostic clarity in PET/CT exams, particularly when aiming for shorter scan times or reduced radiation doses. It improves image quality and lesion conspicuousness in oncological studies, offering the potential for reduced FDG injection doses (e.g., 25% generally, and up to 50% for overweight/obese patients) without compromising quantitative SUV and TBR values. HYPER DLR aids in the detection, localization, evaluation, diagnosis, staging, and monitoring of various conditions across oncology, cardiology, neurology, and infection/inflammation, and can support radiotherapy treatment planning.
Integration Options
HYPER DLR is deployed as an upgrade to all installed United Imaging U.S. molecular imaging systems and is seamlessly embedded across United Imaging's full equipment portfolio, including PET/CT systems like the uMI Panvivo family.
Compliance Status
The uAI HYPER DLR has received FDA 510(k) clearance (K193210) and is one of three commercialized deep learning-based algorithms for PET noise reduction approved by the U.S. FDA. United Imaging, as a company, also has numerous CE-marked applications.
Pricing Model
As part of United Imaging's 'Software Upgrades for Life' program, HYPER DLR is provided as an upgrade to all installed U.S. molecular imaging systems at no additional cost to existing customers with a service contract. The company emphasizes price transparency and simplicity, making the technology available across various price points of their machines.
User Experience
The tool aims to simplify the experience of upgrading system capabilities and enhance the overall experience of owning and using medical imaging equipment. Its AI-enabled workflow is designed to support technologist usability, and clinical evaluations by experienced nuclear medicine physicians have indicated improved image quality.
Support Quality
United Imaging's 'Software Upgrades for Life' program ensures ongoing innovation delivery and support. The company has also expanded its service centers. While not specific to HYPER DLR, United Imaging benefits from enterprise-level support from partners like NVIDIA for its AI-enabled products, indicating a robust support ecosystem.
Implementation Complexity
Described as an upgrade seamlessly embedded into existing United Imaging molecular imaging systems, the implementation for current customers is designed to be straightforward, aligning with the company's goal to reduce complexity.
Evidence Base
Clinical evaluation studies, including one with 52 oncological patients, have demonstrated HYPER DLR's superior performance in noise reduction, improved image quality, liver SNR, and lesion TBR compared to standard OSEM reconstruction. The algorithm was developed based on a U-Net structure, trained with over 300 cases, and validated with 80 cases. It has been implemented in more than 10 sites worldwide.
Physician Tip
Leverage HYPER DLR to optimize PET/CT protocols by considering reduced FDG doses and/or shorter acquisition times, especially for patients who may benefit from lower radiation exposure or improved comfort, such as those with higher BMI. The enhanced image quality, particularly in terms of SNR and CNR, can improve lesion detectability and diagnostic confidence. Integrate this tool into routine oncology, cardiology, and neurology PET/CT interpretations for more consistent and clearer images.
HYPER DLR is a native, embedded AI solution within United Imaging's molecular imaging systems, including the uMI Panvivo PET/CT family. For existing United Imaging customers with a service contract, it is provided as a software upgrade at no additional cost, ensuring seamless integration into their current workflow and equipment.
Details
| Category | Oncology AI, Radiology & Imaging AI |
| Pricing |
Included with system purchase and service contract
|
| Deployment | Software-only device, implemented on previously cleared United Imaging PET/CT systems. |
| Compliance | |
| BAA Available | Unknown AI-estimated |
| HIPAA Compliant | Unknown AI-estimated |
| FDA Status |
1 AI-estimated Received FDA 510(k) clearance (K193210) on August 4, 2020. It is cleared as an image processing function intended to reduce noise of fluorodeoxyglucose (FDG) PET images. |
| Integrations | |
| EHR | Not specified |
| Specialties | Nuclear Medicine, Oncology, Radiology |
Ratings & Reviews
No reviews yet. Be the first to review this tool!
Rate HYPER DLR
Press & Coverage
United Imaging
United Imaging, Debuting at AHRA as a Gold Sponsor, Announces FDA Clearance of Its Latest Deep Learning Reconstruction Technology, uAI HYPER DLR
United Imaging announced FDA clearance for uAI HYPER DLR (Deep Learning Reconstruction), an advanced image processing function designed to reduce noise in FDG PET images. This technology will be deployed to all installed U.S. molecular imaging systems at no additional cost to customers as part of their Software Upgrades for Life program.
2020-08
AuntMinnie
United Imaging gets FDA nod for PET noise reduction
United Imaging received FDA 510(k) clearance for its uAI Hyper Deep-Learning Reconstruction (DLR), which aims to reduce noise in FDG-PET images by using a pretrained neural network. The upgrade will be provided to existing molecular imaging systems in the U.S. at no cost through the company's Software Upgrades for Life program.
2020-08
X-ray Interpreter
HYPER DLR | FDA Radiology AI Device - X-ray Interpreter
HYPER DLR is a software-only device from Shanghai United Imaging Healthcare Co., Ltd. that utilizes a pre-trained convolutional neural network to reduce noise in PET images. It is classified as an Emission Computed Tomography System under regulatory class II.
2020-08
United Imaging
United Imaging Caps Year of Expansion with Major Product Launches at RSNA
United Imaging showcased its next-generation AI technologies at RSNA, including the FDA-cleared uAI HYPER DLR deep learning PET image reconstruction algorithm. This was part of a year of significant product launches and expansion for the company.
2020-11
Journal of Nuclear Medicine
Clinical evaluation on image quality of a deep learning-based denoising algorithm in 18F-FDG PET/CT studies | Journal of Nuclear Medicine
This study investigated the performance of HYPER DLR, a deep learning-based reconstruction algorithm, in 18F-FDG PET/CT studies, finding significant improvements in image quality and lesion conspicuousness compared to standard OSEM reconstruction. The algorithm showed potential for shorter acquisition times or reduced FDG injection doses.
2021-05
PMC (PubMed Central)
Deep learning-assisted PET imaging achieves fast scan/low-dose examination - PMC
HYPER DLR is one of three commercialized deep learning-based algorithms approved by the FDA for reducing noise in PET images, enabling reduced injected activity and shorter acquisition times. Clinical evaluations have shown its superior capability for noise reduction and improved image quality, even in patients with higher BMI.
2022-02
PR Newswire
United Imaging Intelligence Highlights Sustainable AI Solutions at ECR 2025 - PR Newswire
United Imaging Intelligence showcased HYPER DLR for PET/CT systems at ECR 2025, emphasizing its ability to improve image quality by boosting the signal-to-noise ratio by 42%. This is part of their commitment to sustainable AI innovations in healthcare.
2025-02
AuntMinnieEurope
United Imaging showcases AI backbone at ECR 2025 - AuntMinnieEurope
At ECR 2025, United Imaging highlighted HYPER DLR for PET, noting its improvement in image quality by boosting the signal-to-noise ratio by 42%. The company emphasized the integration of AI across its full equipment portfolio.
2025-02
Videos
Product demos, reviews, and walkthroughs for HYPER DLR.
Loading videos...
Frequently Asked Questions
uAI HYPER DLR utilizes deep learning to significantly reduce noise in FDG PET images, improving signal-to-noise and contrast-to-noise ratios. This allows for potential reductions in scanning time and radiation dose while maintaining or enhancing diagnostic image quality.
uAI HYPER DLR has received FDA clearance as an advanced image processing function for PET images. As with any AI in healthcare, adherence to data privacy regulations like HIPAA is crucial, ensuring patient data used for processing remains secure and compliant.
While generally effective, studies suggest uAI HYPER DLR may have limitations for lesions with very low uptake values, where lesion contrast improvement might be less pronounced. Physicians should be aware of these specific contexts when interpreting images.
United Imaging, the developer, has stated that uAI HYPER DLR is deployed to all installed U.S. molecular imaging systems at no additional cost to customers as part of their 'Software Upgrades for Life' program. This aims to provide access to advanced innovations across their product portfolio.
Other FDA-approved deep learning reconstruction algorithms include SubtlePET (Subtle Medical) and AiCE DLR (Canon Medical Systems Corporation). Traditional methods like Ordered Subset Expectation Maximization (OSEM) with Gaussian filters are also widely used, though DLRs often offer superior noise reduction.
uAI HYPER DLR was trained using a large dataset of patient cases and validated by experienced nuclear medicine physicians. Studies have shown its superior capability for noise reduction and improved image quality compared to traditional methods, even with reduced acquisition times or tracer doses.
uAI HYPER DLR's ability to maintain diagnostic image quality with reduced noise allows for the potential to decrease PET/CT scanning time or administered radiotracer dose. This can significantly lower patient radiation exposure without compromising diagnostic confidence.
Suggest an Edit →
| Last Verified: 2026-04-19 | First Added: 2026-04-19