OCT Image Quality Improvement Using Learned Model
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional image processing technologies for OCT images struggle to generate images suitable for diagnosis due to noise and low contrast, making it difficult to accurately ascertain objects, especially when noise reduction and contrast enhancement are inadequate.
Innovation Solution
An image processing apparatus using a learned model, trained with data including images with lower noise and higher contrast, to improve image quality by generating and displaying high-quality images with reduced noise and enhanced contrast, allowing for easier identification of authentic tissue visualization.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If images are acquired multiple times and averaging processing is performed to improve image quality, then image quality is improved, but imaging time increases
Solution Approach 1:
The system performs preliminary actions by acquiring multiple images and performing averaging processing before the actual diagnosis process. This pre-processing step improves image quality in advance, allowing single-image diagnosis later without requiring repeated imaging during the diagnostic phase.
Solution Approach 2:
The system creates multiple copies of the same image through repeated acquisition, then combines these copies through averaging processing. This approach generates a high-quality composite image that represents the true structure more accurately than any single image alone.
2Reliability
If conventional image processing is used to improve image quality, then some image enhancement is achieved, but noise reduction and contrast enhancement remain inadequate for accurate diagnosis
Solution Approach 1:
The system replaces conventional mechanical/image-processing-based enhancement methods with artificial intelligence-based processing. The AI model learns optimal enhancement parameters and transformations from training data, automatically achieving superior noise reduction and contrast enhancement that conventional algorithms cannot accomplish.
Solution Approach 2:
The AI-based processing dynamically adjusts multiple image processing parameters simultaneously (noise filtering strength, contrast enhancement levels, sharpness adjustments) based on the specific characteristics of each input image. This adaptive parameter adjustment achieves optimal diagnostic quality that fixed-parameter conventional methods cannot reach.
3Measurement precision
If high resolution is achieved through conventional methods, then resolution is improved, but noise and low contrast persist making diagnosis difficult
Solution Approach 1:
The AI-based processing converts the harmful effects of noise and low contrast into beneficial outcomes. By learning from training data containing various noise patterns and contrast conditions, the system identifies and suppresses noise while enhancing contrast, effectively transforming image degradation factors into quality improvement opportunities.
Data Source
AI summary
An image processing apparatus includes: an obtaining unit configured to obtain a first image of an eye to be examined; an image quality improving unit configured to generate a second image with at least one of lower noise and higher contrast than the obtained first image using the obtained first image as an input data of a learned model, wherein the learned model has been obtained by using training data including a second image with at least one of lower noise and higher contrast than a first image of an eye to be examined; and a display controlling unit configured to cause the obtained first image and the generated second image to be switched, juxtaposed or superimposed and displayed on a display unit.