| Title: | Holistic Prediction of Cardiovascular Outcomes with Multimodal Vision Transformers |
| Lead Institution: | Technische Universität Dresden |
| Principal investigator: |
Application 92261
WARNING: the interactive features of this website use CSS3, which your browser does not support. To use the full features of this website, please update your browser.
About
We will be investigating if it is possible to develop an artificial intelligence (AI) model using various types of patient data such as blood test results and magnetic resonance imaging (MRI) scans to predict the occurrence of certain heart and circulatory diseases such as high blood pressure, atherosclerosis, heart failure, heart attacks or strokes. Similar projects have been attempted in the past with promising results. We believe that a relatively new type of AI architecture called a vision transformer is well suited for such a task and has great potential when coupled with large, high quality datasets. This is due to greater computational efficiency and greater robustness compared with what is currently the most AI architecture - convolutional neural networks. We will be making use of a wide variety of imaging techniques as well as numerical data from patients such as bloodwork results and blood pressure readings from routine check-ups. As cardiovascular disease is quite a broad field, there are a plethora of potential diseases for us to attempt to predict on the basis of this data. As such, we would require an estimated three years to bring our investigations to completion. The potential benefit from a public health point of view is large. By being able to better identify patients who are at higher or lower risk for certain outcomes, their care can be tailored accordingly. This means greater care focused on mitigating the health impact on those most at risk for given health problems, improving health outcomes, and perhaps reducing unnecessary interventions in those identified as lower risk, resulting in greater quality of life. In addition to patient-centric benefits, these findings may also aid in clinical care and decision-making, potentially reducing cost, time and expertise required. Thus, there is great potential in AI-based assistance, such as what we seek to investigate, in reducing the burden of cardiovascular disease in a wide variety of clinical contexts.1 Publication
| Pub ID | Title | Author(s) | Year | Journal |
|---|---|---|---|---|
| 15944 | Deep learning can predict cardiovascular events from liver imaging | Gregory Patrick Veldhuizen (+11) | 2025 | JHEP Reports |
Enabling scientific discoveries that improve human health