Related Experiment Video
Updated: Jul 1, 2025
A Swin Transformer-Based Model for Thyroid Nodule Detection in Ultrasound Images
Published on: April 21, 2023
Detecting Heart Disease from Multi-View Ultrasound Images via Supervised Attention Multiple Instance Learning
Zhe Huang1, Benjamin S Wessler2, Michael C Hughes1
1Dept. of Computer Science, Tufts University, Medford, MA, USA.
Insights
This study introduces an improved deep learning method for diagnosing aortic stenosis (AS) from echocardiograms. The new approach enhances accuracy and reduces model size by focusing on relevant ultrasound views and using a novel pretraining strategy.
Area of Science:
- Cardiology
- Medical Imaging
- Artificial Intelligence
Background:
- Aortic stenosis (AS) is a serious heart valve condition with significant health consequences.
- Current diagnosis relies on expert interpretation of echocardiograms, which is time-consuming and prone to under-diagnosis.
- Automating AS screening using deep learning requires identifying relevant views and aggregating information for accurate diagnosis.
Purpose of the Study:
- To develop an automated deep learning system for accurate aortic stenosis screening.
- To improve upon existing methods that struggle with image selection and aggregation in echocardiography.
- To enhance the efficiency and accuracy of AS diagnosis through novel artificial intelligence techniques.
Main Methods:
- Developed a novel end-to-end multiple instance learning (MIL) approach for AS detection.
- Implemented a supervised attention mechanism to focus on diagnostically relevant echocardiographic views.
- Utilized a self-supervised pretraining strategy with contrastive learning on the entire study representation.
Main Results:
- The proposed MIL approach demonstrated higher accuracy in AS detection compared to previous methods.
- The new technique effectively identified and utilized relevant echocardiographic views for diagnosis.
- The model achieved improved performance while also reducing overall model size.
Conclusions:
- The novel end-to-end MIL approach with supervised attention and self-supervised pretraining significantly improves automated aortic stenosis detection.
- This method offers a more accurate and efficient alternative to current diagnostic practices for AS.
- The findings suggest a promising direction for AI-driven cardiovascular diagnostics.
Abstract:
Aortic stenosis (AS) is a degenerative valve condition that causes substantial morbidity and mortality. This condition is under-diagnosed and under-treated. In clinical practice, AS is diagnosed with expert review of transthoracic echocardiography, which produces dozens of ultrasound images of the heart. Only some of these views show the aortic valve. To automate screening for AS, deep networks must learn to mimic a human expert's ability to identify views of the aortic valve then aggregate across these relevant images to produce a study-level diagnosis. We find previous approaches to AS detection yield insufficient accuracy due to relying on inflexible averages across images. We further find that off-the-shelf attention-based multiple instance learning (MIL) performs poorly. We contribute a new end-to-end MIL approach with two key methodological innovations. First, a supervised attention technique guides the learned attention mechanism to favor relevant views. Second, a novel self-supervised pretraining strategy applies contrastive learning on the representation of the whole study instead of individual images as commonly done in prior literature. Experiments on an open-access dataset and a temporally-external heldout set show that our approach yields higher accuracy while reducing model size.
