Spatio-Temporal Medical Image Analysis for Motion-Aware IMT Measurement
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing neural networks fail to model motion and orientation changes in sequences of medical images acquired using methods like OCT or ultrasound, leading to inadequate analysis of parameters like Intima-Media Thickness (IMT) due to speckle noise and vessel contortion, especially in UHFUS imaging.
Innovation Solution
A method and system using an artificial neural network that tracks time-varying spatial data and generates spatio-temporal data by correlating visual movements in medical images, incorporating dilated convolutions and residual connections across multiple layers to learn spatio-temporal relationships.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If existing neural networks analyze each image independently, then processing speed is maintained, but motion modeling capability deteriorates
Solution Approach 1:
The patent combines multiple independent image analyses by introducing a motion modeling layer that integrates spatial features from multiple frames. The neural network merges temporal information across frames while maintaining independent processing capabilities, resolving the contradiction between processing speed and motion modeling capability.
Solution Approach 2:
The patent adds a temporal dimension to the analysis by incorporating motion vectors and temporal features into the neural network architecture. This transforms the analysis from purely spatial (2D) to spatio-temporal (3D), enabling motion modeling while preserving processing efficiency through optimized temporal feature extraction.
2Measurement precision
If UHFUS imaging is used to measure IMT, then measurement precision is improved, but speckle noise increases
Solution Approach 1:
The patent introduces motion compensation as an intermediary process between image acquisition and analysis. By calculating motion vectors and applying compensation transformations, the system mediates the effect of speckle noise, allowing precise IMT measurement despite the noisy UHFUS images.
Solution Approach 2:
The patent converts the harmful speckle noise into useful motion information by analyzing speckle patterns across frames to derive motion vectors. The speckle noise, normally an obstacle to measurement, becomes a source of motion data that improves overall measurement accuracy when properly processed.
3Measurement precision
If vessels are imaged at shallow depths, then IMT measurement capability is improved, but vessel contortion increases
Solution Approach 1:
The patent applies dynamic motion compensation that adapts to vessel deformation. Rather than assuming rigid vessel structures, the system dynamically adjusts for contortion by calculating motion vectors that account for pulsation and deformation, maintaining measurement precision despite shape changes.
Solution Approach 2:
The patent performs preliminary motion compensation by calculating and applying motion corrections before IMT measurement. This preliminary action addresses vessel contortion in advance, ensuring that subsequent measurements are not affected by pulsation or deformation.
4Measurement precision
If motion compensation is applied, then analysis accuracy is improved, but processing complexity increases
Solution Approach 1:
The patent segments the motion compensation process into distinct modular steps: feature extraction, motion vector calculation, and compensation application. This segmentation reduces processing complexity by breaking down the complex task into manageable, optimized sub-tasks that can be processed efficiently.
Data Source
AI summary
Provided is a system, method, and computer program product for analyzing spatio-temporal medical images using an artificial neural network. The method includes capturing a series of medical images of a patient, the series of medical images comprising visual movement of at least one entity, tracking time-varying spatial data associated with the at least one entity based on the visual movement, generating spatio-temporal data by correlating the time-varying spatial data with the series of medical images, and analyzing the series of medical images based on an artificial neural network comprising a plurality of layers, one or more layers of the plurality of layers each combining features from at least three different scales, at least one layer of the plurality of layers of the artificial neural network configured to learn spatio-temporal relationships based on the spatio-temporal data.