SuBLIME Algorithm for Automatic Lesion Detection in MRI
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for detecting tissue abnormalities in multiple sclerosis, such as lesion detection in MRI images, are manual, time-consuming, prone to human error, and costly, and lack an efficient way to combine information from multiple imaging modalities.
Innovation Solution
The development of Subtraction-Based Logistic Inference for Modeling and Estimation (SuBLIME), a method that automatically detects tissue abnormalities by normalizing and processing multi-modality MRI data to generate probability maps of lesion incidence, using logistic regression models and incorporating information from FLAIR, PD, T2-weighted, and T1-weighted images.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual slice-by-slice visual inspection by neuroradiologists is used to detect lesions, then detection accuracy can be maintained, but the process becomes very slow and costly
Solution Approach 1:
The patent replaces the mechanical visual inspection process with an automated computer-based image processing system. The system uses signal processing to obtain image data, normalizes it using statistical parameters, processes multiple modalities through a statistical model, and generates probability maps to automatically detect tissue abnormalities, eliminating the need for manual slice-by-slice review while maintaining detection capability
Solution Approach 2:
The system enables self-service by allowing the imaging data to undergo automatic normalization and processing without human intervention. The statistical model automatically analyzes the normalized image data from multiple modalities and generates lesion probability maps independently, making the detection process autonomous and eliminating dependency on manual radiologist review
2Measurement precision
If manual segmentation of serial MRI images is used to compute lesion volume change, then accurate measurement can be achieved, but the process is time consuming and costly
Solution Approach 1:
The patent replaces manual segmentation with automated processing. The system obtains image data from consecutive studies, normalizes it using statistical parameters derived from the image data itself, processes it through a statistical model, and automatically generates probability maps that identify lesion locations and volume changes without requiring time-consuming manual segmentation
Solution Approach 2:
The system performs preliminary normalization of the image data using statistical parameters before processing. This preliminary action prepares the data in advance, ensuring consistent scaling and intensity distribution across multiple modalities and time points, which facilitates accurate automated lesion detection and volume computation without requiring subsequent manual adjustment
3Reliability
If subtraction images are created from multiple imaging modalities, then image quality improves and susceptibility to registration errors decreases, but no method has been developed to combine information from multiple modalities
Solution Approach 1:
The patent merges information from multiple imaging modalities by obtaining image data from different MRI sequences (such as T1-weighted, T2-weighted, FLAIR), normalizing each modality's data using statistical parameters, and then processing them together through a unified statistical model to generate combined probability maps that leverage the strengths of each modality
Solution Approach 2:
The system applies parameter changes by normalizing image data using statistical parameters (such as mean and standard deviation) derived from each modality's image data. This transformation standardizes the data across different modalities, enabling their effective combination while accounting for modality-specific intensity variations and characteristics
4Measurement precision
If two-dimensional T2-weighted subtraction images are used to identify active lesions, then higher number of lesions can be identified with greater observer agreement, but the images are prone to artifacts from misregistration and partial volume effects
Solution Approach 1:
The patent replaces the vulnerable two-dimensional subtraction process with a more robust automated system that processes three-dimensional imaging acquisitions. The system obtains image data from volumetric MRI scans, normalizes them using statistical parameters, and processes them through a statistical model that is less susceptible to registration errors and partial volume effects, reducing artifacts while maintaining lesion identification capability
Data Source
AI summary
A method of automatically detecting tissue abnormalities in images of a region of interest of a subject includes obtaining first image data for the region of interest of the subject, normalizing the first image data based on statistical parameters derived from at least a portion of the first image data to provide first normalized image data, obtaining second image data for the region of interest of the subject, normalizing the second image data based on statistical parameters derived from at least a portion of the second image data to provide second normalized image data, processing the first and second normalized image data to provide resultant image data, and generating a probability map for the region of interest based on the resultant image data and a predefined statistical model. The probability map indicates the probability of at least a portion of an abnormality being present at locations within the region of interest.