Automated Medical Image Abnormality Detection via Sparse Decomposition
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for automated detection of abnormalities in medical images are time-consuming and suffer from high intra- and inter-expert variability, lacking efficient fully automatic solutions that can accurately process large volumes of images with diverse pathological characteristics.
Innovation Solution
The method involves deformable registration of target images to a common template using a subset of normative images, defining a dictionary, and performing sparse decomposition to classify voxels as normal or abnormal, with an iterative process that improves registration and detection accuracy by progressively increasing deformability and excluding estimated abnormal regions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If manual delineation of pathological regions is performed, then detection accuracy is maintained, but processing time increases significantly and intra- and inter-expert variability increases
Solution Approach 1:
The patent introduces a probabilistic atlas as an intermediary framework that combines manual expertise with automated processing. The atlas serves as a mediator between manual delineation methods and fully automated detection, providing a probabilistic model that captures expert knowledge while enabling efficient automated querying and abnormality detection without requiring repeated manual delineation
Solution Approach 2:
The patent performs preliminary manual delineation to create the probabilistic atlas during an offline training phase. This preliminary action captures expert knowledge in advance, allowing subsequent automated detections to leverage this pre-established model without requiring real-time manual intervention, thus resolving the contradiction between speed and accuracy
2Productivity
If fully automatic methods are implemented, then processing time is reduced, but detection accuracy and reliability decrease due to lack of expert guidance
Solution Approach 1:
The probabilistic atlas acts as an intermediary that embeds expert knowledge into the automated system. Rather than using simple automated algorithms, the system uses the atlas as a mediator that provides statistically grounded expectations of normal anatomy, making the automated detection more reliable by incorporating human expertise in a computationally efficient form
Solution Approach 2:
The patent replaces the mechanical system of repeated manual delineation with a computational model based on probability theory. The probabilistic atlas transforms expert knowledge into a mathematical framework that can be efficiently queried by automated algorithms, substituting manual mechanical processes with computational operations that maintain reliability while improving efficiency
3Measurement precision
If high deformability is applied in image registration, then alignment accuracy improves, but the risk of fitting abnormalities increases
Solution Approach 1:
The patent applies local quality control by using the probabilistic atlas to provide location-specific constraints during registration. The atlas provides expected anatomical variations at each location, allowing the registration to be more flexible in regions with high normal variation while maintaining stricter constraints in regions where abnormalities are less likely, thus preventing abnormality fitting while maintaining necessary alignment accuracy
Solution Approach 2:
The patent dynamically adjusts registration parameters based on the probabilistic atlas information. By incorporating atlas-based expectations into the registration cost function, the system changes the effective deformability parameters adaptively - allowing more deformation where anatomical variation is expected and restricting deformation where abnormalities might be fitted, thus resolving the contradiction between alignment accuracy and abnormality fitting risk
Data Source
AI summary
Methods, systems, and computer readable media for automated detection of abnormalities in medical images are disclosed. According to a method for automated abnormality detection, the method includes receiving a target image. The method also includes deformably registering to the target image or to a common template a subset of normative images from a plurality of normative images, wherein the subset of normative images is associated with a normal variation of an anatomical feature. The method further includes defining a dictionary using the subset of normative images. The method also includes decomposing, using sparse decomposition and the dictionary, the target image. The method further includes classifying one or more voxels of the target image as normal or abnormal based on results of the sparse decomposition.