Cone-beam CT Image Enhancement via Generative Adversarial Networks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current Cone-Beam Computed Tomography (CBCT) images suffer from low image quality due to artefacts and noise, which can hinder adaptive treatment planning and diagnosis, and there is a lack of efficient methods to suppress these artefacts effectively.
Innovation Solution
The use of generative adversarial networks (GANs) and cycle generative adversarial networks (CycleGANs) is employed to generate synthetic CT images from CBCT images, leveraging discriminator and generator models to learn from paired CBCT and real CT images, thereby enhancing image quality by reducing or eliminating artefacts.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If conventional CBCT imaging methods are used, then the imaging process is simple and fast, but the image quality is low with significant artefacts and noise
Solution Approach 1:
A generative adversarial network (GAN) is introduced as an intermediary processing system between the CBCT acquisition and final image output. The GAN consists of a generator network that creates enhanced images and a discriminator network that evaluates them, working together to transform low-quality CBCT images into high-quality images that resemble real CT scans while removing artefacts and noise
Solution Approach 2:
The patent replaces traditional image processing methods (filtering, reconstruction algorithms) with an artificial intelligence-based deep learning system. The GAN uses neural network architectures with multiple layers and parameters to perform image enhancement, substituting conventional signal processing mechanics with adaptive learning-based processing
2Manufacturing precision
If multiple artefact suppression methods are applied, then different types of artefacts can be addressed, but the processing becomes complex and no single method suppresses all artefacts
Solution Approach 1:
The GAN framework provides a universal solution that handles multiple types of CBCT artefacts (noise, streaking, beam hardening, scatter) through a single integrated processing system. The generator and discriminator networks work together to simultaneously address various artefact types without requiring separate processing steps for each artefact category
Solution Approach 2:
The patent transforms the image enhancement problem into a parameter optimization problem where the GAN learns optimal transformation parameters during training. The network adjusts internal parameters (weights, biases) to minimize the difference between enhanced CBCT images and reference CT images, automatically adapting to different artefact types and severity levels
3Productivity
If traditional image processing methods are used, then processing is straightforward, but computing processing times are long and resource efficiency is low
Solution Approach 1:
The GAN model is trained in advance using paired CBCT and CT images, learning the complex mapping between low-quality and high-quality images beforehand. During actual CBCT processing, the pre-trained GAN can rapidly enhance images without requiring extensive real-time computation, as the heavy learning work was completed during the preliminary training phase
Data Source
AI summary
Techniques for generating an enhanced cone-beam computed tomography (CBCT) image using a trained model are provided. A CBCT image of a subject is received. a synthetic computed tomography (sCT) image corresponding to the CBCT image is generated, using a generative model. The generative model is trained in a generative adversarial network (GAN). The generative model is further trained to process the CBCT image as an input and provide the sCT image as an output. The sCT image is presented for medical analysis of the subject.