Longitudinal Medical Image Comparison With Finding-Guided Image Pairing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current machine-learning models for medical imaging are limited in their ability to effectively compare longitudinal studies, making it difficult and time-intensive for clinicians to identify subtle changes in medical findings over time, which can lead to inaccurate readings and detrimental patient outcomes.
Innovation Solution
A computer-implemented method that utilizes a machine learning algorithm to generate a change map by selecting image pairs based on medical findings, using a language model to extract findings from reports, and applying image registration to enhance the visualization of changes in longitudinal medical studies.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If machine-learning models are used to compare longitudinal medical images, then the speed of image comparison is improved, but the accuracy of detecting subtle changes deteriorates
Solution Approach 1:
The patent segments the comparison task by first identifying specific regions of interest (ROIs) based on medical findings from prior reports, then focusing the machine-learning comparison only on these segmented regions. This allows the system to maintain high speed while improving accuracy by concentrating computational resources on the most relevant areas where subtle changes are likely to occur.
Solution Approach 2:
The system applies local quality by differentiating between regions requiring detailed analysis and those that can be processed more quickly. By using the machine-learning model to generate change maps across the entire image and then focusing clinical review on specific ROIs identified from prior reports, the system optimizes both speed and detection accuracy for the critical regions.
2Measurement precision
If clinicians manually compare all images in longitudinal studies, then the accuracy of detecting changes is improved, but the time required for assessment deteriorates
Solution Approach 1:
The system performs preliminary actions by automatically extracting medical findings from prior reports using natural language processing, identifying regions of interest, and pre-computing change maps using machine-learning models. This preliminary processing reduces the time clinicians need to manually compare images while maintaining high detection accuracy through focused review of the most relevant changes.
Solution Approach 2:
The patent introduces an intermediary system that includes machine-learning models for generating change maps and natural language processing tools for extracting findings from reports. This intermediary layer processes the bulk of the comparison work automatically, reducing the time clinicians spend on manual comparison while maintaining accuracy through the sophisticated algorithms that detect subtle changes.
3Reliability
If machine-learning models are trained for specific tasks like lesion identification, then the specificity of the model is improved, but its versatility for longitudinal comparison deteriorates
Solution Approach 1:
The patent implements universality by using a machine-learning model that can perform multiple functions: identifying lesions in single images, comparing longitudinal images, detecting subtle changes over time, and generating change maps. The model is trained with diverse datasets that include both single-image lesion detection tasks and longitudinal comparison tasks, enabling it to adapt to different comparison scenarios while maintaining high specificity for lesion identification.
Solution Approach 2:
The system applies dynamics by enabling the machine-learning model to adapt its comparison strategy based on the specific characteristics of each longitudinal study pair. The model can dynamically adjust its analysis focus based on the medical findings extracted from reports, the type of imaging modalities involved, and the specific regions of interest identified, thereby maintaining both specificity and versatility across different longitudinal comparison scenarios.
Data Source
AI summary
A computer-implemented method comprises: obtaining a plurality of prior medical image sequences from a prior study, and a prior report from the prior study; obtaining a plurality of current medical image sequences from a current study; determining at least one medical finding based on the prior report; identifying an image pair based on the at least one medical finding, wherein the image pair includes a prior image from a prior medical image sequence among the prior medical image sequences and a current image from a current medical image sequence among the current medical image sequences; and inputting the image pair to a machine learning algorithm trained to generate a change map indicating at least one difference between the current image and the prior image of the image pair.