Heart disease is notoriously difficult to diagnose, as its symptoms come and go so suddenly. This has increased interest in long-term cardiac imaging and other sophisticated, inclusive, non-invasive and cost-effective monitoring devices like this wearable gadget.
Cardiac imaging is one of the best ways to find and treat heart problems before they get worse.
“The heart suffers from all kinds of different pathologies,” points out co-author Hongjie Hu.
According to the author, the cardiac image will show what is really happening. Whether the volume changes are caused by a strong but normal contraction of the heart chambers or an emergency heart morphological problem, real-time image monitoring of the heart provides a clear picture of the whole situation.
Cardiac imaging is an important clinical tool used to check long-term heart health, find problems as they occur, and help very sick people.
This new wearable heart monitor offers real-time automated analysis of the heart’s pumping activity, even during physical activity, without the need for invasive procedures.
The wearable cardiac monitoring device employs ultrasound to continuously record images of the heart’s four chambers from multiple angles and analyzes a clinically relevant subset of the images in real time using custom artificial intelligence technology. The project builds on what the team has already done to create usable deep-tissue imaging technology.
“The increasing risk of heart disease calls for more advanced and inclusive monitoring procedures,” says lead author Sheng Xu, adding, “By providing patients and clinicians with more complete details, continuous, real-time monitoring of cardiac imaging is poised to optimize and reformulate the paradigm of cardiac diagnostics”.
Existing non-invasive methods, on the other hand, have limited sampling capabilities and provide only limited information. The wearable device created by Xu’s team allows safe, non-invasive and high-quality cardiac imaging, producing images with a high level of contrast, spatial resolution and temporal resolution.
“It also minimizes patient discomfort and overcomes some limitations of non-invasive technologies such as CT and PET that can expose patients to radiation,” adds co-author Hao Huang.
The sensor’s unique design makes it particularly suitable for use with moving bodies. According to Xiaoxiang Gao, a postdoctoral researcher with the Xu group at UC San Diego, the device can be mounted on the chest with little restriction on participants’ mobility, even providing a continuous record of cardiac events before, during and after exercise. . ”
The new system takes information from an adhesive as soft as human skin and designed to adhere well. The patch is approximately the size of a postage stamp, measuring 1.9 cm (W) × 2.2 cm (W) x 0.09 cm (T). It transmits and receives ultrasonic waves that are used to produce a continuous stream of real-time images of the anatomy of the heart. This ultrasonic patch adheres well to human skin even when stretched, and is soft and flexible.
Using ultrasound, the system can view the left ventricle of the heart from two different angles. This makes it possible to obtain more clinically useful images than ever before. As a use case example, the team showed how it is not possible to image the heart while it is exercising with the rigid and heavy equipment used in clinical settings.
Three things determine how the heart works: stroke volume, ejection fraction, and cardiac output (the volume of blood the heart pumps each minute). Stroke volume is the amount of blood the heart pumps out with each beat. The ejection fraction is the percentage of blood pumped out of the left ventricle of the heart with each beat.
The team led by Xu created an algorithm to enable autonomous, real-time processing with the help of AI.
According to Mohan Li, a master’s candidate in the Xu group at UC San Diego, “A deep learning model automatically segments the left ventricle shape from continuous image recording, extracting its volume frame by frame and generating waveforms for measuring stroke volume, cardiac output and ejection fraction”.
According to Ruixiang Qi, a master’s candidate in the Xu group at UC San Diego, the AI component specifically incorporates a deep learning model for image segmentation, an algorithm for calculating cardiac volume, and a data imputation technique.
“We used this machine learning model to calculate heart volume based on the shape and area of the left ventricle segmentation.”
It is the first time that a deep learning model for image segmentation has been implemented in a portable ultrasound system. This makes it possible for the device to provide accurate and continuous waveforms of important cardiac indices in different physical states, such as when the person is standing still or after exercise. This has never been done before.
Thus, this technology can make curves of these three indices continuously and without making any changes to the data. This is possible because the AI part processes the constant stream of images to make numbers and curves.
To make the platform, the team had to deal with some technical issues that required them to make careful decisions. The researchers employed a 1-3 piezoelectric composite bonded with Ag-epoxy backing as the material for transducers in the ultrasonic imager to create the wearable device itself, lowering the risk and increasing the effectiveness compared to previous techniques. Wide-beam composite transmission gave them the best results when deciding how to configure the transducer array for transmission. For machine learning-based image segmentation, they chose the FCN-32 over nine known models, as it had the highest accuracy.
The patch is now connected to a computer, allowing the machine to automatically download data while the patch is still active. A wireless circuit for the patch was created by the team and will be featured in a future post.
Xu and engineer Shu Xiang created Softsonics, a spinoff of UC San Diego, to commercialize this technology.
The work was described in the January 25 issue of the journal Nature.
Image credit: David Baillot/UC San Diego Jacobs School of Engineering