Brightness Adjustment Model for Radiographic Image Halation Prevention
Solution Overview
Problem
Existing brightness adjustment methods for radiographic images struggle to enhance contrast without causing pixel value saturation, particularly in areas critical for diagnosis, and fail to effectively handle exceptional cases like artificial joints, leading to halation or black defects.
Innovation Solution
A trained model using machine learning is developed to output brightness adjustment parameters, where a loss function is configured to bias the contrast reduction when saturation occurs, ensuring the contrast is optimized without exceeding optimal values, and the model is optimized to prevent halation or black defects by weighting errors based on pixel value comparisons.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Illumination intensity
If contrast is increased by brightness adjustment, then visibility in diagnostic areas is improved, but pixel value saturation (halation or black defects) occurs
Solution Approach 1:
The loss function incorporates feedback by monitoring pixel values in the brightness-adjusted image and penalizing cases where saturation occurs. The training model uses this feedback to learn optimal brightness adjustment parameters that enhance contrast while preventing halation or black defects in diagnostic areas.
Solution Approach 2:
The system dynamically adjusts brightness adjustment parameters (such as gamma values, contrast factors, or histogram equalization parameters) based on the input image characteristics and diagnostic area identification. By changing these parameters adaptively rather than applying fixed adjustments, the system optimizes visibility while avoiding pixel saturation.
2Measurement precision
If machine learning model is trained to output optimal brightness adjustment parameter, then contrast enhancement is achieved, but model output varies in direction of increasing contrast beyond optimal value causing saturation
Solution Approach 1:
The loss function is designed asymmetrically to penalize contrast oversaturation more heavily than under-adjustment. When the brightness adjustment parameter causes pixel saturation, the loss function applies a larger penalty, biasing the model output toward safer contrast levels that prevent halation or black defects while still improving visibility.
Solution Approach 2:
The loss function incorporates preliminary anti-action by anticipating potential saturation issues before they occur in actual application. During training, it proactively penalizes parameter outputs that would lead to saturation, teaching the model to avoid these values and inherently preventing the problem before it manifests in diagnostic imaging.
3Productivity
If conventional algorithm-based brightness adjustment is used, then processing speed is maintained, but extraction errors and exceptional cases (artificial joints) cause halation or black defects
Solution Approach 1:
The system replaces conventional algorithm-based brightness adjustment with a machine learning model that has been trained to handle exceptional cases. The trained model substitutes rigid algorithmic rules with learned patterns from diverse training data, including artificial joints and other anomalies, enabling reliable contrast enhancement without halation or black defects while maintaining processing speed.
Data Source
AI summary
This production method of a trained model (10) includes a step of optimizing a training model (11) so that a value of a loss function (A) becomes small. The loss function (A) is configured to output a relatively large value in a case where the contrast of the predetermined area (PA) in a brightness-adjusted image adjusted in brightness becomes high with respect to training data (20) than in a case where the contrast of the predetermined area in the brightness-adjusted image (3) becomes low with respect to the training data.