Deep Learning Noise Reduction for X-ray CT Medical Images
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Medical images from X-ray CT apparatuses face challenges in noise reduction due to the inability to quantify noise using general-purpose image quality evaluation indices, limiting the effectiveness of noise reduction methods, especially when noise is adaptive and purpose-dependent.
Innovation Solution
An image processing apparatus and method utilizing deep learning to construct a learned network with input images including original, noise-reduced, and intermediate images, enabling adaptive noise reduction in medical images by applying a convolutional neural network for improved noise reduction processing.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If deep learning is applied to noise reduction in medical images, then noise reduction effectiveness is improved, but the requirement for large numbers of correct images for training cannot be satisfied
Solution Approach 1:
The patent applies preliminary action by performing noise reduction processing on medical images before using them as training data. The system first reduces noise in original medical images using conventional methods, then uses these pre-processed images as both input and target for deep learning training, eliminating the need for large numbers of manually annotated correct images
Solution Approach 2:
The system applies self-service by using the medical images themselves to generate the training data needed. The deep learning model is trained to map noisy medical images to their denoised versions, where the training targets are derived from the same image dataset through automated noise reduction processing, making the system self-sufficient without requiring external correct image datasets
2Measurement precision
If conventional noise reduction processing is used, then calculation cost is reduced, but noise reduction effectiveness is insufficient for medical images with purpose-dependent noise
Solution Approach 1:
The patent applies parameter changes by training the deep learning model to adapt to different noise characteristics in medical images. The system learns to recognize and remove noise patterns specific to different medical imaging purposes and modalities, achieving superior noise reduction effectiveness compared to conventional fixed-parameter methods while managing computational complexity through efficient network architecture design
Data Source
AI summary
Noise is reduced for a medical image for which noise cannot be quantified by a general-purpose image quality evaluation index. An image processor has a preprocessor that generates input images including an original image and one or more images with reduced noise compared with the original image; and a noise reduction processor outputs an image, which is obtained by reducing noise from the original image based on the input images, by applying a learned network. The learned network used in the noise reduction processor is constructed by performing deep learning using a plurality of learning sets in which one or more of a medical image including noise, a noise-reduced image obtained by performing noise reduction processing on the medical image, and an intermediate image obtained during the noise reduction processing are input images and a correct image is obtained based on the input images an output image.