AI models improve the accuracy of diagnosing coronary artery disease

Several recent discoveries improve the accuracy of diagnosing coronary artery disease and predicting patient risk with the help of artificial intelligence (AI) models developed by scientists at Cedars-Sinai’s Medical Artificial Intelligence Division. indicates that

These advances led by Dr. Piotr Slomka, Director of Innovation in Imaging at Cedars-Sinai and a research scientist in the Artificial Intelligence Division of Medicine and the Smidt Heart Institute, he is making it easier to detect and diagnose one of the most common and deadly heart diseases. I’m trying

Coronary artery disease affects the arteries that supply blood to the heart muscle. If left untreated, it can lead to heart attacks and complications such as arrhythmias and heart failure.

Affected by approximately 16.3 million Americans over the age of 20, the condition is commonly diagnosed using single-photon emission computed tomography (SPECT) and computed tomography (CT) imaging. However, the images produced during scanning are not always legible.

“We continue to show that AI can improve the quality of images and reveal more information. This will enable more accurate disease diagnosis.” AI to improve cardiac imaging A paper on has been published.

Using AI to improve cardiac images

First study published in Journal of Nuclear Medicineuses AI technology for cardiac imaging that helps improve the diagnostic accuracy of SPECT images for coronary artery disease through advanced image enhancement.

Attenuation correction is important in SPECT imaging. This reduces cardiac image artifacts, making them easier to read and more accurate. However, it requires additional CT scans and expensive hybrid SPECT/CT equipment. It is essentially two scanners in one.

CT attenuation correction has been shown to improve the diagnosis of coronary artery disease, but currently only in a few scans due to the additional scan time, radiation, and limited availability of this expensive technique. is running.

To overcome these obstacles, Slomka and his team developed a deep learning model called DeepAC to generate corrected SPECT images without the need for expensive hybrid scanners. These images are generated by AI techniques similar to those used to generate “deepfake” videos and can simulate high quality images acquired by hybrid SPECT/CT scanners.

The team uses uncorrected SPECT images, which are used in most places today, advanced hybrid SPECT/CT images, and new AI-corrected images of unseen data from centers that have never been used in DeepAC. used to compare the diagnostic accuracy of coronary artery disease. training.

They found that AI-produced images were of about the same quality as those obtained with more expensive scanners, allowing similar diagnostic accuracy.

This AI model can generate DeepAC images in fractions of a second with standard computer software and can be easily implemented into clinical workflows as an automated preprocessing step. ”

Dr. Pyotr Slomka, Cedars-Sinai, Director of Imaging Innovation

Prediction of serious adverse cardiac events

In a second study published in Journal of American College of Cardiology: Cardiovascular Imagingthe team demonstrated that deep learning AI can predict major adverse cardiac events such as death and heart attack directly from SPECT images.

Researchers trained an AI model using a large multinational database containing 5 different sites with over 20,000 patient scans. This included images representing the perfusion and motion of each patient’s heart.

The AI ​​model incorporates visual explanations for physicians, highlighting images in areas of high risk of adverse events.

The team then tested the AI ​​model with over 9,000 scans at two separate sites. They found that deep learning models predict patient risk more accurately than software programs currently used in clinics.

“In our initial research, we were able to demonstrate that AI can be used to perform important image corrections without the need for expensive scanners,” says Slomka. “Second, we show that existing images can be used in a better way, predicting a patient’s risk of heart attack or death from the images and highlighting features of the heart that are indicative of that risk to inform clinicians of coronary artery disease.” to provide better information about

“These findings represent a proof-of-principle of how AI can enhance clinical diagnostics,” said Sumeet Chugh, M.D., Ph.D., head of the Artificial Intelligence Division in Medicine. His AI-powered SPECT imaging enhancements could improve the diagnostic accuracy of coronary artery disease, making it much faster and cheaper than current standards. ”

Reduce bias in AI models

A third study published in European Journal of Nuclear Medicine and Molecular Imagingdescribes how to train an AI system to perform well on all applicable populations, not just the population the system was trained on.

Some AI systems are trained using high-risk patient populations, which can cause the system to overestimate disease probabilities. To ensure that the AI ​​model works accurately for all patients and reduces bias, Slomka and his team used a variety of simulated patients to train the AI ​​system. This process, called data augmentation, helps better reflect the mix of patients undergoing imaging studies.

They found that models trained on a balanced mix of patients more accurately predicted the likelihood of coronary artery disease in women and low-risk patients. This could lead to less invasive testing and more accurate diagnosis in women.

The model also leads to lower false positives, suggesting that the system could potentially reduce the number of tests patients undergo to rule out disease.

“This result suggests that it is important to enrich the training data in order to more accurately reflect the population to which AI predictions will be applied in the future,” Slomka said. .

Researchers are currently evaluating these new AI approaches at Cedars-Sinai, investigating how they can be integrated into clinical software and used in standard patient care.

This research was supported in part by the National Heart, Lung, and Blood Institute.