Cardiac Imaging In 2040: The Future Of AI And ML | Circle CVI
July 11, 2022
Cardiac Imaging In 2040: The Future Of AI And ML
Artificial intelligence (AI) and machine learning (ML) have already made their way in cardiac imaging and are making an impact; but how far can these technologies develop in this clinical setting? Will they cement their place in cardiac imaging and continue to be utilized well into 2040? Or will newer technology take its place?
In this article, we will consider the future of cardiac imaging, and the impact made by AI and ML. We will ask whether AI and ML could eventually eliminate the need for radiologists.
Let’s take an overview of the various ways in which AI and ML are being used in cardiac imaging. The benefits of AI and ML are:
- Indication and patient scheduling
- Image reconstruction and improvement in image quality
- Segmentation and quantification
- Classification and reporting
AI and ML algorithms can help clinicians with indication setting and patient scheduling. Research has validated AI-based planning, which offers benefits such as reducing physicians’ workload and cutting the time it takes to generate schedules.
Acquiring good images can sometimes be a challenge, especially when there is complex cardiovascular disease. AI has helped with often cumbersome tasks such as shimming images (the practice of improving field homogeneity in magnetic resonance imaging (MRI) scanning by compensating for magnetic field imbalances). When artifacts overlap with the anatomy of interest, AI can help.
Most of the major imaging vendors are pursuing AI in image reconstruction. There is now deep learning-guided denoising of cardiac images. Accelerated image acquisition can mean noisier images. AI denoising can improve image quality, which can in turn speed up the imaging process, representing a big step forward.
Vendors of imaging hardware and software are starting to implement AI algorithms for segmentation. Automated identification of anatomical structures from coronary computed tomography angiography (CCTA) has been achieved, and the identification of cardiac structures from non-contrast computed tomography (CT) also works reasonably well. These algorithms are complementary to a radiologist’s work, offering new information which is clinically useful for reporting.
Deep learning has been used to estimate the presence of flow-limiting stenosis, using a neural network that compares the texture of the left ventricle of patients with or without significant stenosis – this AI method can classify patients as being positive or negative. This can save patients an unnecessary trip to the cardiac catheterization lab.
By applying neural networks, quantification that radiologists would not usually have time for can be done. This information can be helpful in managing patient prognosis and outcome. A new AI-powered imaging biomarker has improved cardiac risk prediction over and above the current state-of-the-art. Also, AI for automated calcium scoring on radiotherapy planning CT scans has been highlighted as “a fast and low-cost tool to identify patients with breast cancer at increased risk of coronary valve disease”.
AI is excelling in cardiovascular imaging, being used to predict diagnosis and outcomes, and select the most suitable treatment. Compared to older statistical approaches, AI has proven to be a more effective method of handling the abundance of information which is generated by imaging studies, along with data from electronic health records. This gives it the potential to enhance each step in the cardiovascular imaging process; from image acquisition to analysis and reporting. AI can offer faster decision-making and cut out human error.
With cardiac imaging investigations increasing each year, health costs are also on the rise. The potential of AI to increase efficiency and bring down costs is likely to grow even more appealing to clinicians.
Is it possible that AI and ML will replace radiologists in the future? There have been suggestions that AI and ML would be capable of replacing radiologists, being able to tick boxes of reliability and productivity. But rather than replace radiologists altogether, it seems more likely that AI can support radiologists in their roles, optimizing their workflow and assisting in the discovery of genomic markers, and offering faster reporting. With nearly half of radiologists having experienced burnout in their jobs, these benefits are all the more pertinent.
There are potential barriers to the increased implementation of AI in radiology. These include the security risks attached to systems that use large volumes of data, and the need for human supervision in order to ensure that AI is performing tasks efficiently.
However, the benefits certainly seem to outweigh the potential obstacles when it comes to AI’s future in cardiac imaging. AI in radiology can help radiologists to prioritize cases - based on factors such as urgency – and support radiologists in managing their time better. AI can also take some routine tasks off the hands of radiologists, allowing them to concentrate on other issues. This support can help to avoid the burnout and fatigue which is experienced by radiologists, allowing them to enjoy a better work/life balance.
AI offers outstanding opportunities and appears certain to grow within radiology. In the future, AI will be applied to the entire imaging pipeline from patient selection to prognosis. AI can improve cardiovascular imaging and patient outcomes, and radiologists can take a lead role in directing future AI initiatives.
Clinical settings look likely to change as AI becomes more embedded into the routine of cardiac imaging. But radiologists themselves should not be concerned that they will be replaced by technology – AI can help to free radiologists from time-consuming, repetitive work and cut down on wasted time.
Ultimately, the need for human intuition – based on human values such as compassion and empathy – can never be fully replaced by AI. But AI’s potential to assist and enhance the work of a radiologist is likely to be increasingly embraced. For more information, please read our article on AI and ML in arrhythmias and cardiac electrophysiology.