A team of scientists from the Johns Hopkins University has developed a new tool for predicting which individuals affected by one complex inflammatory heart disease are at risk of sudden cardiac arrest. the new approach is the first to create models of patients’ hearts built from multiple images with the power of machine learning (multimedia imaging).
The study was published in the scientific journal Science Advances.
Inflammatory heart disease – that’s what the new technology promises
“This new and robust customized technology has surpassed clinical metrics in forecasting future arrhythmia and could transform the management of patients with cardiac sarcoidosis “, said the senior auteice Natalia Trayanova, Johns Hopkins biomedical engineering professor and co-director of‘Alliance for Cardiovascular Diagnostic and Treatment Innovation.
Physicians currently lack new precision technologies to assess which patients with sarcoidosis cardiac (inflammatory heart disease), a condition that causes inflammation and scarring that can trigger irregular heartbeats, can have a fatal arrhythmia: this lack unfortunately is the consequence of the death of some patients and of the continuous invasive interventions for those who survive. A recent meta-analysis explained that approximately one third of CS patients receive adequate treatment.
“There is an urgent clinical need for better predictive tools”, said Trayanova, who is also a professor at the Johns Hopkins School of Medicine. “Some patients with inflammatory heart disease they die, often in the prime of their lives, while others have an unnecessarily implanted defibrillator and often face complications, including infections, device malfunction and inappropriate shocks, without receiving any real benefit. “
During their research, John Hopkins scientists developed three-dimensional digital models of the heart of 45 patients with CS treated at Johns Hopkins Hospital. To do this, they used a new approach that combines data from two different types of heart scans: contrast-enhanced cardiac MRI, which detects fibrosis or scarring, and PET scans, which detect inflammation.
“These personalized cardiac models are the first of their kind to be created with data from multiple imaging modalities “, said Trayanova, whose lab experimented with mechanistic heart modeling techniques. “The combined effects of fibrosis and inflammation have never previously been represented in cardiac models ”.
The team used computer simulations to apply a series of electrical signals at various locations in each of the models and collected millions of data points that measure each heart’s reaction.
The group of experts exploited some computer simulations to be able to apply a series of electrical signals at various positions in each of the models and collected millions of data points that measure the reaction of each heart:
“We collected extremely dimensional data with the aim of understanding how the various characteristics of scarring and inflammation influenced the heartbeat.“, Said Trayanova.
Scientists, at a later stage, combined mechanistic simulation data, along with additional patient and imaging data, to develop and train an algorithm to predict the likelihood of arrhythmia leading to cardiac arrest: “In a complex disease like CS, with scars and inflammation, learning the results of mechanistic simulation has allowed us to relate them to real world results”, he added Julie Shade, lead author of the study and PhD.
The tool significantly exceeded standard clinical metrics for predicting cardiac arrest in CS patients. To improve the algorithm, the experts developed an intensive cross-validation process, what measures whether the same accuracy can be achieved when different subsets of data are removed, suggesting how the tool might work on future patients. In all, the team conducted 560 cross-validation iterations.
“We were able to estimate the accuracy of the tool for new patients with a confidence of 95%, which means that we were relatively certain that the algorithm would not be distorted by the data it was trained with and would therefore be accurate when applied to new patients “, Shade said.
Finally, the scientists compared their simulations with scans of lesions in the heart of patients who had subsequently undergone a procedure to restore their heartbeats, finding that their insights were consistent with the actual results.
Large clinical trials will be needed, but the researchers hope their tool will transform the management of patients with inflammatory heart disease by reducing the number of unnecessary defibrillator device implants while ensuring the protection of patients at risk of sudden cardiac death.
According to Trayanova, The synergistic use of customized models and machine learning, which has never been used before to address a problem in cardiovascular health care, could also help solve one of the biggest challenges for implementing artificial intelligence in the health sector: lack of data.
Inflammatory heart disease: what happens to patients with covid19
Patients suffering from inflammatory heart disease with the aggravating circumstance of being infected with covid19 deserve special attention. To remedy this, John Hopkins University cardiologists have developed a algorithm which alerts doctors several hours before hospitalized COVID-19 patients experience cardiac arrest or blood clots. (2)
COVID-HEART predictor can predict cardiac arrest in COVID-19 patients with a mean early warning time of 18 hours and predict blood clot formation three days early. This technology was developed with data from 2,178 patients treated in five Johns Hopkins Health System hospitals between March 1 and September 27.
“It’s an early warning system to predict these two outcomes in hospitalized COVID patients in real time“, Said Trayanova. “The continuously updated predictor can help hospitals allocate the appropriate resources and interventions to achieve the best patient outcomes. “
Julie K. Shade, researcher at the Department of Biomedical Engineering, developed the machine learning algorithm with more than 100 clinical data points, demographic information and laboratory results obtained from the JH-CROWN registry that Johns Hopkins established to collect COVID19 data from every individual in the hospital system.
“Over the summer I would see anecdotal reports on Twitter or in pre-presses of some cardiovascular variables in COVID patients that doctors had found could be significant and add them to the model“Said Shade. “It evolved a lot when we learned about COVID. We didn’t know everything that would be important because it is such a new disease ”.
“For example, the team did not predict that the electrocardiogram data would play a critical role in predicting blood clotting. But once added, the ECG data has become one of the most accurate indicators for the condition “, Trayanova explained.
The next step for the researchers will be to develop the best method for installing the technology in hospitals to aid in the care of COVID-19 patients.
“COVID-HEART prediction tool could help in rapid triage of COVID-19 patients in the clinical setting, especially when resources are limited“Said Allison Hays, associate professor of medicine at Johns Hopkins University School of Medicine and clinical collaborator on the project. “This could have implications for the treatment and closer monitoring of COVID-19 patients to help prevent these poor outcomes. “
In Italy, on the other hand, research conducted by Marco Metra of the University of Brescia, director of theCardiology Unit of the ASST-Spedali Civili.
“Our analysis showed that Covid-19 patients with concomitant heart disease have a extremely severe prognosis, significantly worse than the already severe one of non-heart patients with Covid-19 pneumonia. The main causes of mortality were acute respiratory distress syndrome (ARDS), thromboembolic events, including pulmonary embolism, and septic shock “ explains Metra.
“ Studies carried out on Chinese case series had already suggested the greater susceptibility to Covid-19 pneumonia of heart patients and the possibility of heart damage during infection. In this study, for the first time, both clinical features than risk factors for increased mortality of these patients: age, history of heart failure, history of renal failure, diabetes “ continues the scientist.
“The prognostic significance of some simple laboratory parameters is also confirmed, such as creatininemia (blood parameter that indicates renal function, plasma troponin (an important index for heart health), lymphopenia (lack of specific white blood cells)“.
The study involved 99 patients with covid19 pneumonia and previous diagnoses of heart disease: 53 of the subjects observed already had heart problems while 46 did not have concomitant heart disease. Among the heart patients involved, 40% had a history of heart failure, 36% had atrial fibrillation, and 30% had ischemic heart disease; 67 years the average age with 81% of male patients.
Observing all the cases of the study, it was shown that during hospitalization, unfortunately 26% of patients died, 15% had thromboembolic events, 19%, acute respiratory stress syndrome, 6% septic shock. The comparison between cardiopathic individuals and healthy subjects showed the highest mortality of patients with heart disease, 36% against 15% of non-cardiopathic patients with a rate of thromboembolic events and septic shock also higher: 23 against 6%, and 11 % against zero.