Anatomical Constraints for Cardiac MRI Scar Segmentation

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Current deep learning algorithms for segmenting myocardium and scar in late gadolinium enhancement (LGE) cardiac MRI images are prone to anatomical inaccuracies, require manual interventions, and struggle with varying image quality and fibrosis patterns, limiting their ability to generalize across populations and produce consistent results.

Innovation Solution

A fully automated deep learning method involving three stages of neural networks to segment the left ventricle, contour myocardium, and identify scar/fibrosis regions, applying geometric constraints to ensure anatomical accuracy, without requiring manual human intervention.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Extent of automation

If deep learning algorithms are used for myocardium and scar segmentation in LGE cardiac MRI images, then automation and efficiency are improved, but anatomical accuracy and reliability deteriorate due to anatomical inconsistencies and artifacts

Engineering Contradiction:

Improveautomation of segmentationVSAvoidanatomical accuracy

Core Design Contradiction:

Extent of automationVSReliability

Solution Approach 1:

The patent implements feedback by using predicted segmentations from the deep learning model as inputs to the anatomical consistency module, which then generates corrections based on anatomical rules and constraints. These corrections are fed back to refine the final segmentation output, creating a closed-loop system that continuously improves anatomical accuracy while maintaining automation.

Inventive Principle:

Principle #23Feedback

Solution Approach 2:

The patent introduces an anatomical consistency module as an intermediary between the deep learning segmentation model and the final output. This module acts as a mediator that enforces anatomical constraints and corrects inconsistencies without requiring manual intervention, thus preserving automation while improving reliability through structured anatomical knowledge.

Inventive Principle:

Principle #24Intermediary (Mediator)

2Productivity

If existing deep learning algorithms are applied to LGE-CMR images, then processing speed is improved, but measurement precision deteriorates due to sensitivity to varying image quality and fibrosis patterns

Engineering Contradiction:

Improveprocessing speedVSAvoidsegmentation precision

Core Design Contradiction:

ProductivityVSMeasurement precision

Solution Approach 1:

The patent applies parameter changes by dynamically adjusting anatomical constraint parameters based on the specific characteristics of each LGE-CMR image, including fibrosis patterns and image quality metrics. This allows the anatomical consistency module to adapt its enforcement strength and specific constraints to match the input image properties, maintaining precision across diverse imaging conditions without sacrificing processing speed.

Inventive Principle:

Principle #35Parameter changes

3Reliability

If manual interventions are required for landmark specification and boundary labeling, then anatomical accuracy is improved, but ease of operation and productivity deteriorate

Engineering Contradiction:

Improveanatomical accuracyVSAvoidoperational simplicity

Core Design Contradiction:

ReliabilityVSEase of operation

Solution Approach 1:

The patent implements self-service by enabling the system to automatically perform functions previously requiring manual expert intervention. The anatomical consistency module autonomously identifies landmarks, labels boundaries, and corrects anatomical inconsistencies using programmed anatomical knowledge and the input image data alone, eliminating the need for manual operations while maintaining high anatomical accuracy.

Inventive Principle:

Principle #25Self-service

4Adaptability or versatility

If deep learning models are trained on diverse LGE-CMR data, then adaptability is improved, but device complexity increases due to multiple networks and processing stages

Engineering Contradiction:

Improvegeneralization across populationsVSAvoidsystem complexity

Core Design Contradiction:

Adaptability or versatilityVSDevice complexity

Solution Approach 1:

The patent applies segmentation by dividing the overall segmentation task into distinct functional components: a deep learning model for initial segmentation and an anatomical consistency module for correction. This modular segmentation allows each component to be optimized independently and trained on diverse data, improving adaptability while managing complexity through clear functional separation and reusable modules.

Inventive Principle:

Principle #1Segmentation

Data Source

PatentUS20230394670A1Anatomically-informed deep learning on contrast-enhanced cardiac MRI for scar segmentation and clinical feature extraction

Publication Date: 2023.12.07 JOHNS HOPKINS UNIVERSITY

AI summary

Fully automated computer-implemented deep learning techniques of contrast-enhanced cardiac MRI segmentation are provided. The techniques may include providing cardiac MRI data to a first computer-implemented deep learning network trained in order to identify a left ventricle region of interest to generate left ventricle region-of-interest-identified cardiac MRI data. The techniques may also include providing the left ventricle region-of-interest-identified cardiac MRI data to a second computer-implemented deep learning network trained in order to identify myocardium to generate myocardium-identified cardiac MRI data. The techniques may further include providing the myocardium-identified cardiac MRI data to at least one third computer-implemented deep learning network trained to conform data to geometrical anatomical constraints in order to generate anatomical-conforming myocardium-identified cardiac MRI data. The techniques may further include outputting the anatomical-conforming myocardium-identified cardiac MRI data.