Determining the probability of death through artificial intelligence

A research study by Harvard University and Hochschule Stralsund University of Applied Sciences confirms the thesis that individual life expectancy and health can be determined with the help of artificial intelligence.

A joint study by Harvard University and Stralsund University of Applied Sciences to provide new results

A research study by Harvard University and Hochschule Stralsund University of Applied Sciences confirms the thesis that individual life expectancy and health can be determined with the help of artificial intelligence.

In order to exclude suspicion of pneumonia, physicians usually take an X-ray of the chest. Researchers at Harvard University and the Stralsund University of Applied Sciences (in German: Hochschule Stralsund) have now investigated whether the available x-rays can also be used to predict mortality.

The scientists have created an artificial neural network which independently evaluates the image data of x-rays and determines the probability of death. A total of over 55,000 images from two large clinical studies were evaluated, of which about 40,000 images were used for the development of the algorithm and the remaining images for validation. The final algorithm was then used to determine risk classes. The algorithm, which works exclusively on the basis of the image data, takes less than half a second to classify.

"Our results show that artificial intelligence (AI) can be used to extract information about the lifespan and health for routine medical images," explains Prof. Dr. Thomas Mayrhofer from the Stralsund University of Applied Sciences. With regard to the benefit of the study for individual patients, Mayrhofer is certain that "knowledge about the individualized mortality risk can be used to make informed decisions about preventive measures such as lung cancer screening".

Source:
Lu MT, Ivanov A, Mayrhofer T, Hosny A, Aerts HJWL, Hoffmann U.:
"Deep Learning to Assess Long-term Mortality From Chest Radiographs." JAMA Netw Open. Published online July 19, 2019 2(7):e197416. doi:10.1001/jamanetworkopen.2019.7416