Multi-Resolution Model Training for Faster Medical Image Reconstruction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Machine learning models in the medical field face challenges with insufficient training data, leading to hallucinations and prolonged reconstruction times, especially in unsupervised learning processes.
Innovation Solution
A learning apparatus that generates multiple training data sets with varying resolutions, training machine learning models in stages, starting from low resolution and progressively increasing to higher resolutions using transfer learning, to enhance convergence speed and reduce training time.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If unsupervised learning is used for image reconstruction, then training data requirements are reduced, but reconstruction time increases
Solution Approach 1:
The patent segments the training data into multiple datasets with different resolutions (first training data set with lower resolution, second training data set with higher resolution). This allows the model to learn from coarse-to-fine details, improving training efficiency while maintaining reconstruction quality, thus addressing both data quantity constraints and time efficiency
Solution Approach 2:
The patent performs preliminary training on low-resolution data before high-resolution training. This preliminary action allows the model to learn basic patterns and features first, which accelerates subsequent high-resolution training and reduces overall reconstruction time while requiring less high-quality training data
2Reliability
If supervised learning is used with large training data, then model performance improves, but data acquisition becomes difficult
Solution Approach 1:
The patent changes the resolution parameter of training data, creating multiple datasets with different resolutions from the same source data. This allows the system to achieve good model performance without requiring additional large quantities of high-quality annotated data, as the same data can be used at multiple resolution levels
Solution Approach 2:
The patent adds the resolution dimension to the training data. Instead of only using data at one resolution level, the system trains on multiple resolution levels, effectively utilizing the same data in different dimensional forms to improve model reliability without increasing data quantity
3Manufacturing precision
If training is performed on high resolution data directly, then model accuracy improves, but training time increases
Solution Approach 1:
The training process is segmented into stages: first training on low-resolution data to learn basic patterns, then progressively training on higher-resolution data. This segmentation allows the model to build knowledge incrementally, achieving high accuracy without the prohibitive training time that would result from direct high-resolution training
Solution Approach 2:
Low-resolution training serves as a preliminary action that prepares the model for subsequent high-resolution training. This preliminary training establishes foundational features and patterns, making the later high-resolution training more efficient and reducing the overall time required to achieve high model accuracy
Data Source
AI summary
According to one embodiment, a learning apparatus includes processing circuitry. The processing circuitry is configured to obtain a first training data set, the first training data set is based on target data items and having a first resolution. The processing circuitry is configured to train a first machine learning model using the first training data set to generate a first trained model. The processing circuitry is configured to train a second machine learning model using a second training data set and the first trained model to generate a second trained model, the second training data set is based on the target data items and having a second resolution higher than the first resolution.