Automated Liver Disease Classification via Spleen-Liver Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for classifying the severity of liver diseases from non-invasive radiographic images are limited by proprietary and resource-intensive MRI protocols, unreliable readings in obese patients, and variable performance across different liver disease parameters.
Innovation Solution
A method for automated classification of liver disease severity using non-invasive radiographic images, which involves pre-processing images to segment and normalize liver and spleen regions, and inputting extracted parameters into a trained Machine Learning model for classification.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If proprietary MRI protocols (LiverMultiScan) are used to classify liver disease severity, then measurement capability is improved, but device complexity and time consumption increase significantly
Solution Approach 1:
The method segments the complex LiverMultiScan MRI protocol into separate functional components: morphological liver information extraction, spleen information extraction, and machine learning-based classification. This allows the complex measurement task to be divided into manageable steps that can be performed with standard MRI sequences rather than requiring a proprietary multi-parameter protocol.
Solution Approach 2:
The patent introduces a machine learning model as an intermediary that processes extracted morphological and spleen information to classify liver disease severity. This intermediary layer enables accurate classification without requiring direct implementation of complex proprietary MRI measurement protocols, thereby simplifying the overall system while maintaining diagnostic accuracy.
2Measurement precision
If LiverMultiScan MRI protocol is implemented, then liver disease detection capability is improved, but time consumption increases to about 15 minutes
Solution Approach 1:
The method extracts only the essential morphological liver information and spleen information needed for classification, rather than performing the complete LiverMultiScan protocol. This extraction approach retrieves the critical diagnostic data while eliminating time-consuming redundant measurements, reducing scan time from 15 minutes to a fraction of that duration.
Solution Approach 2:
The patent performs preliminary extraction of morphological and spleen information from standard MRI images before applying machine learning classification. By preparing the essential data in advance from routinely acquired images, the method avoids the need for time-consuming proprietary protocol execution while maintaining diagnostic capability.
3Measurement precision
If LiverMultiScan MRI is used for fibrosis assessment, then measurement capability is improved, but reliability decreases in obese patients with significant fat or fluid
Solution Approach 1:
The method uses spleen information as a reference copy or benchmark to assess liver fibrosis. By comparing liver morphological information against spleen characteristics, the system can reliably estimate fibrosis severity even when direct liver measurement is compromised by obesity-related fat or fluid interference, thereby maintaining reliability across diverse patient populations.
4Measurement precision
If LiverMultiScan protocol is applied, then liver disease classification capability is improved, but performance variability increases across different disease parameters
Solution Approach 1:
The patent transforms the classification approach by changing from direct morphological assessment to a machine learning-based system that integrates multiple parameters (liver morphology, spleen characteristics) with learned weightings. This parameter transformation stabilizes performance across different disease parameters by allowing the model to adaptively prioritize relevant features for each specific classification task, reducing variability.
Data Source
AI summary
A method for performing classification of the severity of at least one liver disease from non-invasive radiographic images is disclosed. The method includes: providing radiographic images of slices of the abdomen of a patient; pre-processing the radiographic images by: segmenting liver and spleen, thus achieving a spleen binary mask and a liver binary mask per slice, and normalizing the images with each other, thus achieving normalized radiographic images per slice; for each slice, from the liver binary mask and the normalized radiographic images, extracting a liver parameter; from at least one spleen binary mask, extracting a spleen parameter; and classifying, in function of both parameters and by help of a trained Machine Learning model, the severity of liver disease between one among a group of liver disease at early stage and a group of liver disease at advanced stage.