Camera-Agnostic CAD Detection With Explainable Attention Maps
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing myocardial perfusion imaging techniques, such as SPECT and PET, face challenges in accurately detecting obstructive coronary artery disease due to camera-specific and sex-specific normal limits, leading to difficulties in diagnosis and the lack of explainable deep learning models that provide black-box results.
Innovation Solution
A deep learning model that operates agnostically across different camera types, utilizing polar maps and patient information to generate obstructive CAD scoring data and attention maps, providing explainable insights into the diagnostic process.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If deep learning techniques are used to detect obstructive disease, then diagnostic accuracy can be improved, but the results are presented in a black-box fashion that reduces explainability
Solution Approach 1:
The patent introduces attention maps as an intermediary visual representation that mediates between the deep learning model's internal processing and the physician's interpretation. These attention maps highlight the specific regions of the polar maps that influenced the diagnostic decision, providing explainability without compromising the accuracy of the deep learning model.
2Measurement precision
If camera-specific normal limits and separate models are used to accurately detect obstructive disease, then detection accuracy can be improved, but device complexity and the need for multiple models increases
Solution Approach 1:
The patent develops a universal deep learning model that can process and interpret polar maps from different camera types (SPECT and PET) using a single unified approach. This universal model eliminates the need for camera-specific separate models while maintaining accurate detection across different imaging modalities.
3Ease of operation
If traditional MPI techniques with camera-specific normal limits are used, then diagnostic process can be simplified, but accuracy in detecting obstructive disease decreases
Solution Approach 1:
The deep learning model performs self-adjustment and adaptation by automatically learning the appropriate diagnostic criteria from training data across multiple camera types. This self-service capability allows the model to maintain simplicity in operation while achieving high accuracy, as it adapts to different camera characteristics without requiring manual calibration or separate models.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
Enables accurate, fast, and explainable detection of obstructive coronary artery disease, reducing the need for camera-specific normal limits and enhancing clinical interpretation efficiency.
Implementation Method 1
it has been found that deep learning can be utilized to detect obstructive disease
Data Source
AI summary
A deep learning model for the detection of obstructive coronary artery disease (CAD) can take a set of polar maps and patient information as input, then output obstructive CAD scoring data, such as probabilities of obstructive CAD associated with various cardiac territories, as well as an attention map and a CAD scoring map. The model can operate agnostic of camera type used to capture the set of polar maps. The attention map indicates regions of the polar maps important to the deep learning process for that particular set of polar maps. The attention map and obstructive CAD scoring data can be used to generate a CAD scoring map showing CAD probability by segment on a standard 17-segment model of a left ventricle. The attention map and/or CAD scoring map can act as easily explainable tools for interpreting the results of a myocardial perfusion imaging study.