Japanese researchers have developed a predictive method that combines meteorological and seasonal data to know when the population at risk is most likely to suffer a heart attack
Japanese scientists have developed a predictive method that detects when cardiac arrest cases are most likely to occur in a population and can improve patient prognosis. The results presented by the team show 90% success rate.
According to researchers in a study published in the prestigious medical journal Heart, cardiac arrests that occur outside the medical supervision of a hospital have a lower survival rate. Furthermore, in previous studies, it has also been seen that this risk is aggravated when certain weather conditions occur.
Japanese researchers have developed an artificial intelligence algorithm that dives into the enormous amount of meteorological data and can identify patterns that other statistical systems miss.
- Brief Anger Hampers Blood Vessel Function Leading to Increased Risk of Heart Disease and Stroke – New Study
- New Blood Test Pinpoints Future Stroke Risk – Study Identifies Inflammatory Molecules as Key Biomarker
- Enceladus: A Potential Haven for Extraterrestrial Life in its Hidden Ocean Depths
- New Experiment: Dark Matter Is Not As ‘DARK’ As All We Think
- Scientists in Fear of This New Predator From Red Sea Eating Native Species in Mediterranean
In their study, they used meteorological data (temperature, relative humidity, rainfall, snowfall, cloud cover, wind speed, and atmospheric pressure readings) and temporal data (year, season, day of the week, time of day, and holidays) to train artificial intelligence.
How this predictive system works
The experiment is based on a sample of 1,299,784 cases collected by the Japanese out-of-hospital cardiac arrest registry between 2005 and 2013. The algorithm was applied to 525,374, using meteorological or temporal data, or both. These results were compared with 135,678 cases that occurred in 2014-15 to see if the model was capable of predicting the number of daily cardiac arrests in other years.
To verify the accuracy of the algorithm, the researchers also developed a map with the location of all out-of-hospital cardiac arrests that occurred in the Japanese city of Kobe between January 2016 and December 2018.
According to the results provided by the team, the risk of suffering a cardiac arrest is higher on Sundays, Mondays, holidays and when temperatures drop sharply within the same day or between consecutive days.
Although this study was only compared with data from out-of-hospital cardiac arrest in the city of Kobe, the researchers believe that it is broadly generalizable to any population in a developed country.
They further add that:
“The methods developed in this study serve as an example for a new predictive analysis model that could be applied to other clinical outcomes of interest related to life-threatening acute cardiovascular diseases.”
90% accuracy
The study has been reviewed by several scientists, including Dr. David Foster Gaieski, a professor at the Sidney Kimmel Medical College at Thomas Jefferson University in Philadelphia. Foster Gaieski has found that the model was 90% accurate in predicting the number of cardiac arrests that occurred during the study week (24 predicted arrests, 27 observed arrests). Being able to correctly identify the number in four of the seven districts and being mistaken for a strike in each of the three remaining districts.
Although for Foster Gaieski the system still needs a larger sample that collects data from other places with greater geographic and cultural diversity, the finding seems very promising.
“Knowing what the weather is likely to be in the next week can generate cardiovascular emergency alerts for people at risk, notifying the elderly and other patients of upcoming periods of greatest danger,” he says.
Foster Gaieski also suggests that this alert system could help emergency medical services and cardiac catheterization personnel to be prepared when cases are expected to increase.
Artificial doctors
Although this technology is still in its infancy, artificial intelligence is a very useful tool in medicine due to its efficiency in image diagnosis and also due to its enormous capacity for probabilistic calculation.
In addition to predicting a patient’s risk of contracting potential diseases, it can help, for example, to more accurately assess the survival chances of patients.
The team at Medical Brain, Google’s artificial intelligence research unit for predicting diseases and their symptoms, long ago introduced an algorithm capable of making predictions about the probability of survival among hospital patients.
In an article published in the journal Nature, they showed the results of their algorithm’s predictions in patients from two different hospitals. Google’s medical brain was 95% accurate in the first hospital and 93% in the second.
According to the study, these models are more effective at predicting outcomes and metrics than traditional methods.
“These models outperformed traditional predictive models in clinical use in all cases. We believe this approach can be used to create accurate and scalable predictions for a variety of clinical scenarios,” the researchers stated.
Imaging of these digital brains is also well advanced, and several studies show that its results are better than those obtained by human physicians. These algorithms have shown excellent precision and sensitivity for the detection of small radiographic anomalies that are very useful in the diagnosis of different ailments such as cancer or coronary heart disease.
Even so, there is still time for us to see this technology implanted in our hospitals. As this Lancet article indicates, unless AI algorithms are trained to distinguish between benign abnormalities and clinically significant lesions, improved image sensitivity could lead to an increase in false positives as well as from puzzling scenarios in which the AI findings are not associated with the results.
In the end, the success of this technology depends on the quality of the data that is used for its training and many times that information contains biases that can cause inequalities in the result without our realizing it.
- Brief Anger Hampers Blood Vessel Function Leading to Increased Risk of Heart Disease and Stroke – New Study
- New Blood Test Pinpoints Future Stroke Risk – Study Identifies Inflammatory Molecules as Key Biomarker
- Enceladus: A Potential Haven for Extraterrestrial Life in its Hidden Ocean Depths
- New Experiment: Dark Matter Is Not As ‘DARK’ As All We Think
- Scientists in Fear of This New Predator From Red Sea Eating Native Species in Mediterranean
Dr. Irina Grigorescu from King’s College London, analyzes the benefits and limitations of artificial intelligence applied to medicine in this article.
For Grigorescu “Machine-learning techniques currently depend on high-quality datasets. Any bias introduced in the data could therefore be used as a “shortcut” for the algorithm to exploit. It’s a problem that extends beyond medical imaging. Such biases will have to be removed from our training datasets for the AI to become accurate enough for use in the real world.”
Image Credit: Getty