AI Cardiac Motion Classification Using Neural Network Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional methods for classifying cardiac motion based on Cardiac Magnetic Resonance (CMR) cine imaging are highly subjective and time-consuming, requiring 30-60 minutes for evaluation.
Innovation Solution
A computer-implemented method using artificial intelligence (AI) that segments CMR image data through trained neural networks to classify cardiac motion by distinguishing between heart muscle and surrounding tissue, extracting movement features, and associating them with pre-determined motion classes, significantly reducing classification time.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional visual analysis methods are used for cardiac motion classification, then the assessment can be performed with simple equipment and procedures, but the classification time is excessively long (30-60 minutes) and subjectivity reduces reliability
Solution Approach 1:
The patent replaces the mechanical visual analysis system (human observer examining CMR images) with an automated AI-based image processing system. The AI model automatically segments myocardium, extracts motion features, and classifies wall motion abnormalities, eliminating manual visualization and reducing classification time from 30-60 minutes to significantly shorter duration while improving reliability through objective, repeatable automated analysis
Solution Approach 2:
The system enables self-service by allowing the CMR imaging data to automatically undergo segmentation, feature extraction, and classification without requiring continuous human intervention. The AI model processes the images independently, generating wall motion classification results autonomously, thereby reducing dependency on manual analysis and minimizing classification time
2Productivity
If automated AI-based classification is implemented, then classification speed and reliability are improved, but the device complexity and computational requirements increase
Solution Approach 1:
The patent applies segmentation by dividing the complex classification task into distinct modular components: (1) myocardium segmentation to isolate heart muscle tissue, (2) motion feature extraction to identify relevant dynamics, and (3) classification to categorize wall motion abnormalities. This modular AI architecture improves processing efficiency and productivity while managing system complexity through organized, independent functional modules
Solution Approach 2:
The system utilizes parameter changes by transforming raw CMR image data into extracted motion features that capture essential dynamics. The AI model processes temporal and spatial parameters from sequential images, converting complex image sequences into simplified motion metrics that accelerate classification speed while maintaining diagnostic accuracy
3Loss of information
If detailed segmental wall motion assessment is performed manually, then comprehensive diagnostic information is obtained, but the time required for evaluation increases significantly
Solution Approach 1:
The patent implements continuous useful action by processing the entire CMR image sequence through the AI pipeline without interruption. The system continuously segments myocardium across all time frames, extracts motion features throughout the cardiac cycle, and performs classification on the complete dataset, ensuring comprehensive diagnostic information is captured while reducing total evaluation time through uninterrupted automated processing
Solution Approach 2:
The system applies preliminary action by pre-processing the CMR images through automated segmentation and motion feature extraction before classification. This preparatory processing organizes the data structure and identifies key motion patterns in advance, enabling faster and more accurate wall motion assessment while preserving complete diagnostic information about segmental abnormalities
Data Source
AI summary
A computer-implemented method for providing a cardiac motion classification based on Cardiac Magnetic Resonance (CMR) image data, wherein the CMR image data comprise a plurality of image frames, I(x, y, z, t), acquired for respective two-dimensional slices in at least one longitudinal direction, z, of the heart and for a plurality of times, t, the method including: a myocardium segmentation step of inputting the plurality of image frames into two or more trained neural networks, applying the trained neural networks in parallel, and fusing an output of each of the trained neural networks into a single output indicating a segmentation, for each of the plurality of image frames, between a first portion indicating muscle tissue of the heart and a second portion indicating surrounding tissue of the heart muscle, and determining a corresponding mask of muscle tissue for the first portion; a slice classification step of assigning each of the plurality of image frames in each slice, z, to an anatomic layer of the heart; a movement feature extraction and classification step of, for each of the masks and the corresponding anatomic layers, extracting a movement feature of the heart and classifying the movement feature into one of a number of pre-determined movement features; an associating step of associating the classified movement feature with the corresponding layer for the cardiac motion classification.