Machine Learning Medical Image Processing for Lower-Radiation Imaging
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing medical imaging technologies require significant investment or expose subjects to risks to achieve high-quality images suitable for diagnosis, often resulting in images with noise or low contrast that hinder accurate identification of lesions.
Innovation Solution
A medical image processing apparatus using a machine learning engine to improve image quality by generating a composite image from a first and second image, enhancing image quality and displaying it on a display unit.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a high-performance imaging apparatus is purchased to obtain high image quality, then image quality is improved, but investment cost increases
Solution Approach 1:
The patent uses a machine learning engine to generate a synthetic high-quality image (second image) that copies the essential diagnostic features of images obtained from high-performance imaging apparatuses. This allows low-performance apparatuses to produce diagnosis-quality images without purchasing expensive equipment, resolving the contradiction between image quality and investment cost.
2Measurement precision
If the amount of radiation is increased to obtain an image with less noise, then noise is reduced, but the subject is exposed to more radiation
Solution Approach 1:
The patent converts the harmful effect of noise in low-radiation images into a benefit by using a machine learning engine trained to recognize and reconstruct true image features from noisy inputs. The engine learns from training data to distinguish signal from noise, allowing noise reduction without increasing radiation exposure.
3Measurement precision
If a contrast medium is used to obtain an image in which a site is enhanced, then site enhancement is improved, but the risk of side effects increases
Solution Approach 1:
The machine learning engine generates a synthetic enhanced image that copies the contrast enhancement effect without using actual contrast media. By training on pairs of non-enhanced and contrast-enhanced images, the engine learns to predict enhancement patterns, achieving site enhancement without the harmful side effects of contrast agents.
4Measurement precision
If the imaging region is widened or spatial resolution is increased, then imaging quality is improved, but imaging time becomes longer
Solution Approach 1:
The machine learning engine performs preliminary processing by generating a high-resolution synthetic image from a lower-resolution input image. This pre-processing step allows subsequent diagnostic work to proceed with high-resolution images without requiring time-consuming high-resolution scanning in the first place, resolving the contradiction between spatial resolution and imaging time.
5Measurement precision
If imaging is performed multiple times to obtain a high quality image, then image quality is improved, but the time required for imaging increases
Solution Approach 1:
Instead of acquiring multiple images and performing time-consuming averaging or reconstruction, the machine learning engine generates a single high-quality image by copying and enhancing features from a single input image. This synthetic image generation process is computationally efficient and produces diagnosis-quality images without requiring multiple acquisitions.
Data Source
AI summary
A medical image processing apparatus includes: an obtaining unit configured to obtain a first image that is a medical image of a predetermined site of a subject; an image quality improving unit configured to generate, from the first image, a second image in which image quality is improved compared to the first image, using an image quality improving engine including a machine learning engine; and a display controlling unit configured to cause a composite image obtained by combining the first image and the second image according to a ratio obtained using information relating to at least a partial region in at least one of the first image and the second image to be displayed on a display unit.