Longevity tables are interesting but have little relevance to individuals except as they are members of a group. It’s interesting to view longevity data from various countries and then make an effort to understand why one country varies from another. In Australia, for example, babies born in 2015 are expected to live 80.4 years if they’re male and 84.5 years if they’re female; the combined predicted longevity is 81.5, according to the World Health Organization. For comparison, the world average for both sexes is 70.5 years. Other interesting comparisons are New Zealand with 80 for males and 84 for female, the U.K. 79 and 83, and the U.S. 76 and 81. This data doesn’t answer the questions individuals have about how long they can expect to live. Tools that consist of questionnaires about lifestyle, habits, and family members are also interesting and reinforce the value of clean living, but the numbers are still based on big data.
Researchers at the University of Adelaide‘s Schools of Public Health and School of Computer Science used artificial intelligence to analyze CAT scans of 48 patients’ chests to try to predict which ones would die within five years. In a study published in Nature‘s journal, Scientific Reports, the AI predictions were accurate 69% of the time, which is comparable to the predictions by physicians. According to the researchers, their study was the first using AI with medical images. While the algorithms were not more accurate than humans in this test, their equivalency means there’s a potential to turn the laborious and time-intensive task of studying images over to machines so the physicians can focus more on giving care. It’s also significant that the machine learning reached the accuracy level of humans with just 48 samples. Because machine learning never stops, exposing the algorithms to more scans could theoretically result in greater accuracy. In analyzing the results, the scientists could not pinpoint exactly what factors in images made a difference to the algorithms but noted greater accuracy with emphysema and congestive heart failure, both of which can be severe chronic diseases.
The group plans next to present tens of thousands of patient images for machine analysis. The way forward also includes using scans of additional parts of the body to make predictions, not just about life or death, but about impending medical events such as heart attacks.
