Classifier-Based Vascular Function Determination From Perfusion Imaging
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for determining vascular functions in perfusion imaging, such as arterial input and venous output functions, are technically challenging and require tedious manual annotation, limiting their applicability in clinical practice and constraining the selection of regions of interest, leading to suboptimal results.
Innovation Solution
A computer-implemented method using a trained classifier, preferably a convolutional neural network, to determine vascular functions by optimizing voxel-wise weights based on the similarity between predicted and ground truth vascular functions, allowing for automated and accurate determination without spatially constraining the selection to specific regions of interest.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual annotation is used to determine vascular functions, then accuracy can be maintained, but the process becomes tedious and time-consuming
Solution Approach 1:
The system enables automatic self-determination of vascular functions through a classifier that processes perfusion imaging sequences independently, eliminating the need for manual annotation while maintaining accuracy. The classifier automatically identifies and weights voxel time series to determine arterial input and venous output functions.
Solution Approach 2:
The patent replaces the manual mechanical annotation process with an automated classifier system that uses machine learning algorithms to determine vascular functions. This substitution eliminates human intervention while preserving measurement precision through computational analysis of perfusion imaging data.
2Ease of operation
If regions of interest are spatially constrained for vascular function determination, then the selection process is simplified, but the accuracy of vascular function determination deteriorates
Solution Approach 1:
The system dynamically determines regions of interest without pre-specified spatial constraints. The classifier automatically identifies appropriate voxels based on their temporal signal characteristics and weighting, allowing the region selection to adapt to the specific perfusion imaging sequence rather than being fixed by anatomical assumptions.
Solution Approach 2:
The patent changes the approach from spatial parameter-based selection to temporal parameter-based selection. Instead of constraining regions by anatomical location, the system uses temporal signal characteristics (time series profiles) to identify and weight appropriate voxels, thereby improving accuracy while maintaining operational simplicity.
3Extent of automation
If a supervised segmentation approach is used to determine vascular functions, then automation is achieved, but the complexity of the system increases
Solution Approach 1:
The classifier system is designed to perform multiple functions: it automatically determines both arterial input and venous output functions, identifies appropriate weighting for different voxel time series, and handles various perfusion imaging scenarios. This multi-functionality achieves comprehensive automation without requiring separate specialized systems for each task.
Solution Approach 2:
The patent introduces a classifier as an intermediary component that mediates between the raw perfusion imaging data and the final vascular function determination. This intermediary layer processes the imaging sequences, applies temporal weighting, and produces the vascular functions, thereby achieving automation while managing system complexity through a single integrated component.
Data Source
AI summary
A computer-implemented method for determining a vascular function of a perfusion imaging sequence, includes the steps of: (i) receiving a perfusion imaging sequence including a voxel time series for a plurality of voxels; (ii) applying a trained classifier on the perfusion imaging sequence for receiving voxel-wise weights; (iii) receiving voxel-wise weights from the classifier; and (iv) determining the vascular function as the weighted sum of the voxel time series; wherein the classifier is trained by optimizing over the similarity between a predicted vascular function and a ground truth vascular function using a set of examples.