Useful AI technology to measure salt intake only with photos of food before and after meals
Jul 18, 2024
|
Excessive salt intake is known to increase the risk of cardiovascular diseases such as high blood pressure and myocardial infarction, and to cause chronic diseases throughout the body such as renal failure, stomach cancer, and osteoporosis. The World Health Organization (WHO) recommends such salt intake at 2,000 mg per day, but caution is needed because the average daily intake in Korea is still 1.6 times the appropriate standard.
In order to properly manage these salt intake, first of all, it is necessary to know exactly how much salt an individual consumes per day. However, there are many practical restrictions on accurately recording and evaluating the list of foods and each intake for each meal in daily life.
Until now, the '24-hour urine sodium test' is considered the most accurate when hospitalized due to diseases that require limiting sodium intake, such as kidney disease. However, even this is cumbersome to store and test every time you pee several times a day, so a simpler and more applicable method is needed in everyday life.
In response, a research team led by professors Ryu Ji-won and Kim Hye-won of the Bundang Seoul National University Hospital and Kim Se-joong of the Department of Nephrology conducted a study to verify the usefulness of the technology to estimate sodium intake only with food photos, paying attention to artificial intelligence, which has recently been rapidly developing.
The AI used in the study is a model that uses 'YOLO (You Only Look Once) v4' architecture that detects food areas, MST++ to classify food types, ResNet-101 artificial neural network model, and hyperspectral imaging technology to measure food volume.
The research team took pictures of food before and after meals consumed by patients admitted to Seoul National University Bundang Hospital and made AI calculate sodium intake, which was compared and analyzed with the 24-hour urine sodium results.
As a result, it was confirmed that considering variables such as gender, age, kidney function, and diuretics in the AI analysis results, values close to those of the 24-hour urine sodium test can be obtained. Furthermore, we have also succeeded in deriving a formula to predict actual 24-hour urine sodium test results with only the estimated glomerular filtration rate (eGFR) that evaluates sodium intake and kidney function measured by AI.
For example, patients without diuretics can indirectly calculate the results of a 24-hour urine sodium test by adding 22.102 times the estimated glomerular filtration rate to 53.5% of the sodium intake measured by AI.
The results of this study confirmed the possibility of a more convenient AI sodium intake measurement technology for hospitalized patients, and it is expected that it can be widely used not only in clinical settings but also in daily life through advancement in the future.
Professor Ryu Ji-won said, "Because you only need to take photos of food before and after meals with a smartphone application, it is a much easier way to self-assessment records and surveys. If the estimated glomerular filtration rate is used, it can predict urine sodium levels for 24 hours, which is highly likely to be used for inpatients."
Professor Kim Se-joong said, `High salt intake can lead to elevated blood pressure throughout the body, which can damage the glomerulus of the kidneys and surrounding blood vessels"Management is important in daily life because chronic high blood pressure can lead to a vicious cycle that worsens further," he said. "AI sodium measurement technology can be a good means."
Meanwhile, the results of this study were recently published in the international journal of healthcare 'JMIR Formative Research'.
|
bellho@sportschosun.com
