MLP Medical Image Processing With Patch Shifting
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional deep learning techniques for medical imaging, such as MRI, suffer from issues like blurriness, loss of details, and high noise levels in generated images.
Innovation Solution
Utilizing a multi-layer perceptron (MLP) neural network to process medical images through iterative patch splitting, shifting, and concatenation, along with padding and training to enhance image quality, particularly for under-sampled MRI data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If conventional deep learning techniques are used for medical image processing, then the processing speed and computational efficiency are improved, but the image quality deteriorates with blurriness, loss of details, and high noise levels
Solution Approach 1:
The patent divides the input image into multiple patches and processes them separately through the neural network. This segmentation allows the model to focus on local details while maintaining computational efficiency, resolving the contradiction between processing speed and image quality by handling images in manageable segments rather than as a whole
Solution Approach 2:
The patent introduces a new architectural dimension by using multi-layer perceptron networks with skip connections that operate across different patch segments. This dimensional approach allows global information to be integrated without sacrificing local detail processing, thereby improving image quality while maintaining processing efficiency
2Device complexity
If conventional deep learning approaches process medical images, then the computational complexity is reduced, but the detail preservation and noise reduction capabilities worsen
Solution Approach 1:
By segmenting the image into patches, the computational complexity is distributed across multiple smaller processing units. Each patch can be processed with simpler operations while the overall detail preservation is achieved through the aggregation of processed patches with skip connections maintaining global context
Solution Approach 2:
The skip connections act as intermediaries that transmit original patch information directly to later processing stages. This intermediary mechanism allows detailed local information to be preserved and integrated with processed features, improving detail preservation without significantly increasing computational complexity
3Manufacturing precision
If global characteristics are used for image processing, then the noise reduction and detail recovery are improved, but the processing time and computational resources increase
Solution Approach 1:
The patent processes global characteristics by segmenting the image into patches that can be handled in parallel. This segmentation enables the network to capture global information through the aggregation of local patch features while reducing processing time through parallel computation across multiple patches
Solution Approach 2:
The patent applies partial processing by focusing computational resources on processing representative patches rather than the entire image at full resolution. This partial action approach allows global characteristics to be captured efficiently while maintaining detail recovery capabilities in the reconstructed full image
Data Source
AI summary
Described herein are systems, methods, and instrumentalities associated with using a multi-layer perceptron (MLP) neural network to process medical images of an anatomical structure. The processing may include padding an input image in accordance with the training of the MLP neural network, splitting the input image (e.g., the padded input image) into patches of a same size, and processing the patches through the MLP neural network over one or more iterations. During an iteration of the processing, the patches may be processed separately and re-combined into an intermediate image before the intermediate image is shifted to concatenate portions of the image that are derived from different patches. This way, global features of the anatomical structure may be learned and used to improve the quality of the image generated by the MLP neural network, without incurring significant computation or memory costs.