Could YOU tell how old these people are? AI can: Tech is able to pinpoint if someone is in their 20s, 60s or 80s by looking at an X-ray of their chest
- Artificial intelligence is so accurate that it can determine age based on a breast scan
- The tool, developed by scientists in Japan, was fed images from x-rays
Asked to look at these four x-rays, the average person will see no difference.
But AI can do it.
In fact, research shows that artificial intelligence is now so accurate that it can determine someone’s exact age based solely on a breast scan.
Developed by scientists in Japan, the tool obtained images from thousands of real X-rays and was repeatedly tested and fine-tuned.
Test results showed that the system could accurately predict age, with an average age of just three years.
Between 2008 and 2021, chest X-rays were collected from 36,000 healthy patients. Scans from three locations were used to train and tune the AI bot. Results, published in The Lancet Healthy Longevity, showed that the technology’s estimate of a person’s age correlated “very strongly” with their actual age. Salient graphs show that people aged 21 to 40 were much colder in the lower lung area than people aged 81 to 100. This is emphasized by the faded blue spots. Those aged 81 to 100 show much more red areas around the heart, indicating that it is much hotter. The chest x-rays also show that the chest cavity is typically much larger for people aged 21 to 40 compared to those aged 81 to 100. The images appear blurry because they were created with downsampled images. Pictured: X-ray scans (top) and saliency charts (bottom)
Experts also said that when it is wrong and overstated, it is likely a sign that the patient has some disease.
Scientists at Osaka Metropolitan University say their findings represent a “leap in medical imaging” and could pave the way for earlier detection and treatment of diseases such as high blood pressure, lung disease and kidney failure.
Between 2008 and 2021, chest X-rays were collected from 36,000 healthy patients.
Scans from three sites were used to train and tune the AI bot.
Resultspublished in The Lancet Healthy Longevity, showed that the tech’s estimate of someone’s age correlated “very strongly” with their actual age.
Researchers also showed chest scans of another 34,000 patients with a range of underlying conditions, such as asthma, diabetes and liver disease.
When the AI suspected someone was older than they actually were, it was a sign that they were more likely to have diseases like chronic obstructive pulmonary disease (COPD) and osteoporosis.
All of these conditions cause changes in the chest, such as the blood vessels, lungs, bones and heart, that can be detected on an X-ray, the researchers said.
They looked at saliency maps — heat maps that highlight areas on the scan that the AI used to make a decision about a patient’s age.
These suggest that the bot focused on the mediastinum – the space that separates the lungs and holds the heart and vital arteries – to determine a patient’s age.
The scientists said looking at this part of the scan can reveal how much calcium has accumulated around the organs, with high levels being a sign of disease.
Yasuhito Mitsuyama, lead author of the study, said: “Our results suggest that chest radiography-based apparent age may accurately reflect health status beyond chronological age.
‘We want to further develop and apply this research to estimate the severity of chronic diseases, predict life expectancy and predict possible surgical complications.’