MRI Background Noise Reduction via Neural Network Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional MRI background noise reduction methods using fixed intensity thresholds often result in loss of anatomical information and poor contrast due to masking of anatomical regions with lower intensity values, leading to 'holes' in the images and increased background noise.
Innovation Solution
A method involving segmentation of MR images or parametric maps into foreground and background using a trained neural network, with selective application of an intensity threshold only to the background, allowing higher threshold values without masking foreground pixels, thereby reducing background noise while preserving anatomical details.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Object-affected harmful factors
If a fixed intensity threshold is applied to suppress background noise, then background noise is reduced, but anatomical regions with lower intensity values are masked creating holes in the images
Solution Approach 1:
The patent segments the MR image into foreground (anatomical regions) and background regions using a neural network classifier. This segmentation allows different processing to be applied to each region: the background receives aggressive noise suppression while the foreground is preserved with its original intensity values, preventing information loss.
Solution Approach 2:
The patent applies different quality characteristics to different parts of the image. The background region receives strong noise suppression with high thresholding, while the foreground anatomical regions maintain their original signal quality and intensity distribution, ensuring that each region is processed according to its specific requirements.
2Object-affected harmful factors
If a higher intensity threshold is applied to reduce background noise, then background noise suppression is improved, but anatomical details with lower intensity are lost
Solution Approach 1:
By segmenting the image into foreground and background, the patent enables the application of a higher intensity threshold to the background without affecting the foreground. The neural network classification ensures that anatomical structures are identified and protected from aggressive thresholding that would otherwise be necessary to suppress background noise.
Solution Approach 2:
The patent applies local quality enhancement by preserving the original signal characteristics in anatomical regions while applying strong noise suppression to the background. This localized processing maintains anatomical detail precision in critical areas while achieving aggressive noise reduction elsewhere.
3Device complexity
If a fixed intensity threshold is used for background suppression, then processing simplicity is maintained, but image quality deteriorates due to information loss
Solution Approach 1:
The patent replaces the simple mechanical thresholding operation with a neural network-based classification system. This substitution increases processing complexity but enables superior image quality by intelligently distinguishing between background noise and anatomical signals, allowing for more sophisticated and effective noise suppression.
Solution Approach 2:
The patent changes the processing parameters dynamically based on the classified region. Instead of using a fixed threshold for the entire image, the system adjusts the effective threshold applied to each pixel based on its classification as foreground or background, optimizing image quality for each region independently.
Data Source
AI summary
Methods and systems are provided for reducing background noise in magnetic resonance (MR) image and parametric map visualization using segmentation and intensity thresholding. An example method includes segmenting the MR image or parametric map into foreground which includes a region of anatomy of interest and background which is outside of the region of anatomy of interest, applying an intensity threshold to the background and not applying the intensity threshold to the foreground of the MR image or parametric map to produce a noise reduced MR image or noise reduced parametric map, and displaying the noise reduced MR image or noise reduced parametric map via a display device.