Parameterized AI Medical Image Enhancement With Runtime Appearance Editing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing AI-based medical image enhancement techniques exacerbate visual artifacts and are computationally expensive, with predefined visual properties that cannot be adjusted at runtime, limiting their applicability.
Innovation Solution
An AI transformation model with a neural network that predicts transformation function parameters, allowing real-time editable output image appearance control through a graphical user interface, using a simple CNN-based encoder and regression layers, and a transformation module to apply adjustable transformation functions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If existing AI techniques directly infer transformed medical images, then image enhancement quality is improved, but visual artifacts are exacerbated and computational cost increases
Solution Approach 1:
The patent introduces a transformation function as an intermediary between the input medical image and the output enhanced image. Instead of directly inferring the transformed image, the AI model first predicts parameters of a transformation function (such as brightness, contrast, sharpness parameters), which then applies controlled modifications to the original image. This intermediary approach prevents direct pixel manipulation that causes artifacts while maintaining enhancement quality.
Solution Approach 2:
The patent transforms the image enhancement problem from direct pixel transformation to parameter-based transformation. The AI model predicts parameters (e.g., brightness factor, contrast factor, sharpness factor) that control the transformation function. By adjusting these parameters rather than directly transforming pixels, the system achieves high-quality enhancement without the harmful artifacts that result from direct pixel manipulation.
2Manufacturing precision
If existing AI techniques use complex model design for high-resolution images, then image enhancement quality is improved, but computational expense increases
Solution Approach 1:
The patent segments the image enhancement task into two distinct components: (1) predicting transformation parameters using a relatively simple AI model, and (2) applying the transformation function to generate the enhanced image. This segmentation allows using a computationally efficient model for parameter prediction while maintaining high output quality through the transformation function, avoiding the need for complex end-to-end models that process high-resolution images directly.
Solution Approach 2:
The patent uses a transformation function that can be applied repeatedly and efficiently to generate enhanced images. Once the transformation parameters are predicted, the same transformation function is used to transform the image, which is computationally much cheaper than running complex AI models. This copying approach allows high-quality enhancement with reduced computational expense.
3Ease of manufacture
If existing AI techniques train models to generate output images with predefined visual properties, then model training is simplified, but runtime adjustability is lost
Solution Approach 1:
The patent makes the transformation function dynamic and adjustable at runtime. Instead of training the model to output fixed visual properties, the system predicts parameters for a transformation function that can be adjusted by users after the model has made its prediction. The transformation function accepts parameter inputs that control the degree and type of transformation applied, allowing real-time adjustment of visual properties without retraining the model.
Solution Approach 2:
The patent performs preliminary action by training the AI model to predict transformation parameters rather than fixed image outputs. During training, the model learns to predict parameters (brightness, contrast, sharpness) that control the transformation function. This preliminary parameter prediction enables flexible runtime adjustment, as users can modify these parameters to achieve desired visual properties without requiring model retraining.
Data Source
AI summary
Medical image enhancement processing techniques are described that employ an artificial intelligence (AI) model with editable output image appearance control. An example, computer-implemented method comprises generating a transformed version of a medical image via execution of an AI transformation model on the medical image, wherein the AI transformation model comprises a neural network that predicts values of parameters of a transformation function based on processing the medical image, or a down sampled version thereof, via the neural network, and a transformation module that generates the transformation function using the values and applies the transformation function to the medical image, resulting in generation of the transformed version of the medical image. The AI transformation model comprises an editable output image appearance control functionality that enables a user to control and edit the transformation operations performed by the transformation function on the medical image in association with viewing the results in real-time.