A patient-specific artificial neural network training system and method
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing artificial neural networks struggle to accurately diagnose patient-specific conditions due to scarce data, particularly in rare cases, leading to challenges in incorporating patient-specific knowledge and achieving reliable medical diagnoses.
Innovation Solution
A patient-specific training system and method that separates image data into high-confidence and low-confidence datasets, using a classification module to refine the neural network with the high-confidence dataset, enhancing its adaptability to individual patient characteristics.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a neural network is trained with general population data, then it can handle common cases well, but it fails to accurately diagnose patient-specific conditions due to scarce data for rare cases and individual patients
Solution Approach 1:
The patent segments the training data into two distinct datasets: a first dataset containing images with high classification confidence and a second dataset containing images with low classification confidence. This segmentation allows the system to separately handle general cases (first dataset) and patient-specific/rare cases (second dataset), resolving the contradiction between general diagnostic accuracy and patient-specific adaptability.
Solution Approach 2:
The patent applies preliminary classification to all input images before training the neural network. By pre-classifying images and separating them into high-confidence and low-confidence datasets, the system prepares tailored training data that enables the neural network to learn both general patterns and patient-specific characteristics, thereby improving both diagnostic accuracy and adaptability.
2Adaptability or versatility
If the neural network is trained with scarce patient-specific data, then it can adapt to individual patient characteristics, but the number of training samples becomes too small to extract meaningful information
Solution Approach 1:
The patent merges the first dataset (high-confidence images) and second dataset (low-confidence images) into a unified training structure. The neural network processes both datasets, allowing it to learn from the abundant general case data while also capturing patient-specific patterns from the scarce rare case data, thus overcoming the limitation of small sample sizes for individual patients.
Solution Approach 2:
The patent introduces an intermediary classification step that acts as a mediator between the input images and the neural network training process. This intermediary classification separates images into confidence-based datasets, enabling the system to effectively utilize scarce patient-specific data by contextualizing it within a broader framework of general medical data.
3Productivity
If the system processes all image data uniformly, then it maintains simple processing flow, but it cannot distinguish between common and rare cases leading to missed diagnostic opportunities
Solution Approach 1:
The patent segments the image processing flow into distinct stages: preliminary classification to separate high-confidence and low-confidence images, followed by targeted neural network processing. This segmentation maintains processing efficiency by handling common cases through the first dataset while dedicating attention to rare cases through the second dataset, thereby improving diagnostic reliability without sacrificing productivity.
Data Source
AI summary
The invention relates to a training system and a training method for training an artificial neural network, in particular for medical applications, with patient-specific features. The patient-specific artificial neural network training system comprises an input interface, configured to receive image data from a patient, a computing device, further comprising a classification module, a separator module and a training module, and an output interface, configured to output a diagnosis signal, wherein the classification module is configured to acquire as input the received image data and generate a first classification signal for each image of the image data, wherein the separator module is configured to, based on the first classification signal, separate the received image data in at least a first dataset and a second dataset according to a reliability criterion, and wherein the training module is configured to use the first dataset or only a part thereof as a training dataset to train an artificial neural network.