尊龙凯时 - 人生就是搏!

Transforming image reconstruction, DELTA is a deep learning algorithm that advances medical imaging. This advancement enhances imaging capabilities, benefiting patients and the radiologists who care for them. With this feature integrated into our products, it enhances the accessibility and affordability of high-quality imaging, enabling more patients to receive advanced diagnostics, regardless of their location. Discover the future of CT imaging with DELTA, where precision meets innovation.

Two Sides of the Same Coin: Clarity and Care

In CT imaging, achieving a balance between clarity and care is a significant challenge. While low radiation doses enhance patient safety, it is crucial to maintain sufficient image quality for accurate diagnoses. Various techniques have been developed to reduce radiation exposure, such as adjusting tube current and voltage, but these often compromise image quality.

Reconstruction algorithms are vital for effectively balancing radiation dose and patient care. Over several generations, these algorithms have significantly evolved, with AI-based reconstruction now recognized as the most promising approach. Current AI methods excel in image quality, dose reduction, and reconstruction speed, establishing them as the preferred choice in modern CT imaging.

DELTA

Deep learning trained algorithm (DELTA), leverages an extensive training dataset and a sophisticated 3D neural network to deliver sharper, more detailed images at lower radiation dose, setting a new standard in imaging excellence.

80%**

Up to 80% radiation dose reduction at the same low contrast detectability (LCD).

98%**

Up to 98% image noise reduction at the same dose.

155%**

Up to 155% LCD improvement at the same dose.


Conventional 2D neural networks can only process one single image at a time, which may affect the image continuity and lead to missing structural details around lesions. In contrast, 3D neural networks support volumetric data input, maintaining data continuity while improving reconstruction speed, accuracy, and reliability.

Building a 3D Neural Network for CT Imaging

Conventional 2D neural networks can only process one single image at a time, which may affect the image continuity and lead to missing structural details around lesions. In contrast, 3D neural networks support volumetric data input, maintaining data continuity while improving reconstruction speed, accuracy, and reliability.

Million-level Dataset

Leveraging a million-level data training set across diverse clinical scenarios, DELTA delivers whole-body low dose imaging with fine anatomical details.
＞1 million

training dataset

＞5 million

model training epochs

＞6 million

parameters

During the training of neural networks, data propagation and correction are crucial factors for algorithm accuracy. Traditional network structures involve single-layer data propagation, whereas specialized network designs enable dense connections and efficient data propagation between different layers of the network, preserving image details even at low radiation doses.

Specialized Network Design

During the training of neural networks, data propagation and correction are crucial factors for algorithm accuracy. Traditional network structures involve single-layer data propagation, whereas specialized network designs enable dense connections and efficient data propagation between different layers of the network, preserving image details even at low radiation doses.

In instances of severe image quality degradation in low-dose CT (LDCT), traditional reconstruction algorithms often lead to tissue blurring and residual artifacts. DELTA's innovative 3D neural network effectively addresses these challenges. Evaluations conducted using a public LDCT database demonstrate that DELTA significantly enhances imaging quality across various dose levels, ensuring clearer and more accurate diagnostic outcomes.

—— < Domain Progressive 3D Residual Convolution Network to Improve Low Dose CT Imaging >
IEEE Trans Med Imaging (2019)

DELTA was compared with the Hybrid Iterative Reconstruction (HIR) algorithm. The results clearly demonstrated that CT images reconstructed with DELTA achieved significant noise reduction while maintaining diagnostic image quality. Even when the radiation dose was reduced by half, DELTA produced images with less noise than HIR, as confirmed by both subjective and objective analyses.


——< Deep learning trained algorithm maintains the quality of half-dose contrast-enhanced liver computed tomography images: Comparison with hybrid iterative reconstruction Study for the application of deep learning noise reduction technology in low dose >
European Journal of Radiology(2021)