Artificial intelligence deep learning system can automatically measure coronary artery calcium from routine CT scans and predict cardiovascular events like heart attacks

Coronary artery calcification — the buildup of calcified plaque in the walls of the heart’s arteries — is an important predictor of adverse cardiovascular events like heart attacks. Coronary calcium can be detected by computed tomography (CT) scans, but quantifying the amount of plaque requires radiological expertise, time and specialized equipment. In practice, even though chest CT scans are fairly common, calcium score CTs are not. Investigators from Brigham and Women’s Hospital’s Artificial Intelligence in Medicine (AIM) Program and the Massachusetts General Hospital’s Cardiovascular Imaging Research Center (CIRC) teamed up to develop and evaluate a deep learning system that may help change this. The system automatically measures coronary artery calcium from CT scans to help physicians and patients make more informed decisions about cardiovascular prevention. The team validated the system using data from more than 20,000 individuals with promising results. Their findings are published in Nature Communications.

“Coronary artery calcium information could be available for almost every patient who gets a chest CT scan, but it isn’t quantified simply because it takes too much time to do this for every patient,” said corresponding author Hugo Aerts, PhD,  director of the Artificial Intelligence in Medicine (AIM) Program at the Brigham and Harvard Medical School. “We’ve developed an algorithm that can identify high-risk individuals in an automated manner.”

Working with colleagues, lead author Roman Zeleznik, MSc, a data scientist in AIM, developed the deep learning system described in the paper to automatically and accurately predict cardiovascular events by scoring coronary calcium. While the tool is currently only for research purposes, Zeleznik and co-authors have made it open source and freely available for anyone to use.

“In theory, the deep learning system does a lot of what a human would do to quantify calcium,” said Zeleznik. “Our paper shows that it may be possible to do this in an automated fashion.”

The team began by training the deep learning system on data from the Framingham Heart Study (FHS), a long-term asymptomatic community cohort study. Framingham participants received dedicated calcium scoring CT scans, which were manually scored by expert human readers and used to train the deep learning system. The deep learning system was then applied to three additional study cohorts, which included heavy smokers having lung cancer screening CT (NLST: National Lung Screening Trial), patients with stable chest pain having cardiac CT (PROMISE: the Prospective Multicenter Imaging Study for Evaluation of Chest Pain), and patients with acute chest pain having cardiac CT (ROMICAT-II: the Rule Out Myocardial Infarction using Computer Assisted Tomography trial). All told, the team validated the deep learning system in over 20,000 individuals.

Udo Hoffmann, MD, director of CIRC@MGH who is the principal investigator of CT imaging in the FHS, PROMISE and ROMICAT, emphasized that one of the unique aspects of this study is the inclusion of three National Heart, Lung, and Blood Institute–funded high-quality image and outcome trials that strengthen the generalizability of these results to clinical settings.

The automated calcium scores from the deep learning system highly correlated with the manual calcium scores from human experts. The automated scores also independently predicted who would go on to have a major adverse cardiovascular event like a heart attack.

The coronary artery calcium score plays an important role in current guidelines for who should take a statin to prevent heart attacks. “This is an opportunity for us to get additional value from these chest CTs using AI,” said co-author Michael Lu, MD, MPH, director of artificial intelligence at MGH’s Cardiovascular Imaging Research Center. “The coronary artery calcium score can help patients and physicians make informed, personalized decisions about whether to take a statin. From a clinical perspective, our long-term goal is to implement this deep learning system in electronic health records, to automatically identify the patients at high risk.”

Funding for this work was provided by the National Institutes of Health (NIH-USA U24CA194354, NIHUSA U01CA190234, NIH-USA U01CA209414, NIH-USA R35CA22052, 5R01-HL109711, NIH/NHLBI 5K24HL113128, NIH/NHLBI 5T32HL076136, NIH/NHLBI 5U01HL123339), the European Union—European Research Council (866504), as well as the German Research Foundation (1438/1-1, 6405/2-1), American Heart Association Institute for Precision Cardiovascular Medicine (18UNPG34030172), Fulbright Visiting Researcher Grant (E0583118), and a Rosztoczy Foundation Grant.

Paper cited: Zeleznik et al. “Deep convolutional neural networks to predict cardiovascular risk from computed tomography” Nature Communications DOI: 10.1038/s41467-021-20966-2

MVA delivered by skin scarification provoked a potent immune response in mice, suggesting a new strategy for lung-targeted T cell vaccines for respiratory illnesses such as COVID-19

Among infectious diseases that have caused pandemics and epidemics, smallpox stands out as a success story. Smallpox vaccination led to the disease’s eradication in the twentieth century. Until very recently, smallpox vaccine was delivered using a technique known as skin scarification (s.s.), in which the skin is repeatedly scratched with a needle before a solution of the vaccine is applied. Almost all other vaccines today are delivered via intramuscular injection, with a needle going directly into the muscle, or through subcutaneous injection to the layer of tissue beneath the skin. But Thomas Kupper, MD, chair of the Department of Dermatology, and colleagues, had reason to suspect that vaccines delivered by skin scarification may offer better protection against respiratory diseases. In a study published in Npj vaccines, Kupper and co-authors present results from preclinical studies suggesting skin scarification may help generate lung T cells and provide protection against infectious diseases, with implications for prevention of COVID-19.

“We have known for years that this technique was a good way to generate T cells that would home to the skin, but our study shows that skin scarification is also an effective way to generate T cells that home to the lungs,” said Kupper. “Vaccine development today is focused on selecting the best antigen(s) for T cells and B cells. But for a vaccine to work to its full potential, it also needs to direct T cells to where they are needed most. For respiratory pathogens, that means getting T cells to the lungs.”

Historically, smallpox vaccines used live vaccinia virus (VACV). More recently, the Food and Drug Administration has approved the use of modified vaccinia Ankara (MVA), a modern alternative that lacks about 10 percent of the parent genome and cannot replicate in human cells, thus avoiding the serious side effects seen with VACV. MVA, as a smallpox vaccine, is injected subcutaneously.

Kupper and colleagues set out to determine if the skin scarification route of immunization with MVA could provoke a more effective T cell response than other routes of immunization. The team inoculated mice using either skin scarification, intramuscular, subcutaneous, or intradermal injection. Skin scarification generated more T cells, produced greater numbers of lung-specific T cells and provided superior protection against lethal viral doses than the others.

“We used to think that lung-homing T cells could only be generated by direct lung infection, but here we find overlap between T cells appearing after lung infection and T cells generated through skin scarification,” said Kupper.

The authors note that their work is preclinical — until clinical trials are conducted in humans, it’s unknown if the phenomenon seen in the mouse model can be replicated in people. But the work has spurred the Kupper lab to explore the potential for using the MVA vector and skin scarification technique to develop more powerful — and, potentially universal — vaccines against other infectious illnesses such as influenza and coronaviruses.

“We have known for a while that you can program T cells to go where you want them to go in the body — if you want protective T cells in the lungs, this is one way to achieve that. It is a serendipitous finding, but it seems to work very well,” said Kupper.

This work was supported by the National Institutes of Health/NIAID (R01 AI127654) and the National Institutes of Health/NIAMS (R01 AR065807). Kupper and a co-author are inventors on a patent relevant to this study (US8691502B2). The patent is now owned by Pellis Therapeutics, a biotech company specializing in vaccines, in which Kupper and a co-author have equity.

Kupper has received funding from the NIAID to study MVA s.s. immunization with constructs encoding SARS CoV2 proteins, including Spike as well as other conserved T cell antigens (R01 AI127654).

Paper cited: Pan Y et al. “Epicutaneous immunization with modified vaccinia Ankara viral vectors generates superior T cell immunity against a respiratory viral challenge” Npj vaccines DOI: 10.1038/s41541-020-00265-5

Findings on a novel anti-inflammatory pathway may guide researchers toward innovative multiple sclerosis or brain tumor treatments

Astrocytes are the most abundant type of cells within the central nervous system (CNS), but they remain poorly characterized. Researchers have long assumed that astrocytes’ primary function is to provide nutrients and support for the brain’s more closely scrutinized nerve cells; over the years, however, increasing evidence has shown that astrocytes can also actively promote neurodegeneration, inflammation, and neurological diseases. Now, a team led by researchers from Brigham and Women’s Hospital, has shown that a specific astrocyte sub-population can do the opposite, instead serving a protective, anti-inflammatory function within the brain based on signals regulated by the bacteria that reside in the gut. Findings on the new anti-inflammation pathway are published in Nature.

“Over the years, many labs, including mine, have identified important roles for astrocytes in promoting neurological diseases,” said corresponding author Francisco Quintana, PhD, of the Ann Romney Center for Neurologic Diseases at the Brigham. “This is the first case in which we’re showing that at least a subset of these cells (astrocytes) can prevent inflammation. The reason we haven’t seen this before was because we were studying these cells as if they were uniform, or one single cell type. But now we have the resolution to see the differences between these cells.”

The researchers used refined gene- and protein-analysis tools to identify the novel astrocyte subset. The astrocyte population resides close to the meninges (the membrane enclosing the brain) and expresses a protein called LAMP1, along with a protein called TRAIL, which can induce the death of other cells. These features help the LAMP1⁺TRAIL⁺ astrocytes limit CNS inflammation by inducing cell death in T-cells that promote inflammation.

To determine what mechanism controls LAMP1⁺TRAIL⁺ astrocytes in the brain, the researchers performed a series of tests using the gene-editing tool CRISPR-Cas9. They found that a particular signaling molecule, called interferon-gamma, regulates TRAIL expression. Moreover, they found that the gut microbiome induces the expression of interferon-gamma in cells that circulate through the body and ultimately reach the meninges, where they can promote astrocyte anti-inflammatory activities.

Understanding the mechanisms driving the anti-inflammatory functions of LAMP1⁺TRAIL⁺ astrocytes could enable researchers to develop therapeutic approaches to combat neurological diseases, like multiple sclerosis. For example, they are exploring probiotic candidates that can be used to regulate the astrocytes’ anti-inflammatory activity. Additionally, the research team’s more recent data indicates that certain brain tumors exploit this pathway to evade the body’s immune response. The investigators are therefore developing cancer immunotherapies to retaliate against the tumors’ attacks.

“Finding microbiome-controlled anti-inflammatory subsets of astrocytes is an important advance in our understanding of CNS inflammation and its regulation,” Quintana said. “This is a very novel mechanism by which the gut controls inflammation in the brain. It guides new therapies for neurological diseases, and we believe that this mechanism could contribute to the pathogenesis of brain tumors.”

Quintana’s lab identified the only other subset of astrocyte known to be regulated by the gut microbiome in 2016, but the investigators believe that there are likely others. “It’s becoming clear that the gut flora are important in many diseases,” he said. “We’re lucky that we’ve been leading the charge to identify different subsets of astrocytes and the mechanisms that control them. We have a list of other populations of astrocytes, and we’re working to see how the gut flora may control them.”

This work was supported the National Institutes of Health (NS102807, ES02530, ES029136, AI126880, AI149699, DP2AT009499, 1K99NS114111, F32NS101790, and R01AI130019), National MS Society (RG4111A1 and JF2161-A-5), the International Progressive MS Alliance (PA-1604-08459), a Chan-Zuckerberg Initiative Ben Barres Early Career award, the Burroughs Wellcome fund, the Canada Institute of Health Research, Canadian Foundation for Innovation, Dana-Farber Cancer Institute (T32CA207201), Program in Interdisciplinary Neuroscience at BWH, and the Women’s Brain Initiative at BWH, an Alfonso Martín Escudero Foundation postdoctoral fellowship, the European Molecular Biology Organization (ALTF 610-2017), the FAPESP BEPE (#2019/13731-0).

Paper cited: Sanmarco, LM, et al. “Gut-licensed IFN-g⁺ NK cells drive LAMP1+TRAIL+ anti-inflammatory astrocytes” Nature DOI: 10.1038/s41586-020-03116-4