Artificial Neural Network for Medical Image Classification and Generation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current medical imaging systems rely on manual adjustments and preset processing steps to enhance images, which may not consistently highlight clinically important features effectively, limiting the accuracy and clarity of automated AI analysis.
Innovation Solution
A method and system that utilize an artificial neural network (ANN) trained on data pairs of raw sensory signals and processed images, with a global loss function that minimizes classification prediction errors and image similarity losses, enabling the ANN to predict classifications and generate enhanced images similar to those produced by medical imaging systems.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual image processing adjustments are applied to enhance images for human review, then image quality and clinical feature visibility are improved, but processing time and operational complexity increase
Solution Approach 1:
The patent replaces manual mechanical image processing adjustments with an automated neural network system. The neural network automatically performs image enhancement and classification tasks that previously required manual radiologist intervention, thereby improving image quality while reducing processing time and operational complexity.
Solution Approach 2:
The neural network system performs self-service by automatically enhancing images and generating classifications without requiring manual intervention. The system independently processes raw sensory signals, applies learned transformations, and outputs enhanced images with clinical features highlighted, eliminating the need for time-consuming manual adjustments.
2Productivity
If preset processing steps are used to generate images, then processing speed is improved, but image accuracy and clinical relevance deteriorate
Solution Approach 1:
The patent implements dynamic image processing where the neural network adapts its processing based on the specific characteristics of each input image. Rather than applying fixed preset processing steps, the network dynamically adjusts enhancement parameters and feature highlighting based on learned patterns from training data, thereby maintaining both high processing speed and image accuracy.
Solution Approach 2:
The neural network changes processing parameters dynamically based on the input image characteristics. The system learns optimal processing parameters from training data and applies appropriate transformations to each new image, ensuring high accuracy while maintaining efficient processing speed through parameter optimization rather than brute-force manual adjustment.
3Productivity
If AI algorithms are used to automate image analysis, then productivity and standardization are improved, but dependence on training data quality and model generalization becomes a limiting factor
Solution Approach 1:
The patent incorporates feedback mechanisms where the neural network's classifications and image enhancements are evaluated against ground truth data during training. The system learns from feedback signals (loss functions) to continuously improve its performance, ensuring high reliability and accuracy. This feedback-driven approach allows the system to maintain high productivity while achieving reliable diagnostic accuracy through iterative learning and validation.
Data Source
AI summary
A method which includes: Obtaining a training set which comprises: multiple data pairs each comprising: (i) a raw sensory signal acquired by a medical imaging system, and (ii) a processed image generated by the medical imaging system from the raw sensory signal; and a classification label for each of the data pairs. Based on the training set, training an artificial neural network (ANN), wherein the training comprises minimizing a global loss which is a weighted sum of: a loss between the classification labels and classification predictions by the ANN, and a similarity loss between the processed images and images generated by an intermediate layer of the ANN. The training is such that the trained ANN is configured, for a new raw sensory signal: to predict a new classification, and to generate a new image by the intermediate layer of the ANN.