according to study subordinate Massachusetts Institute of Technology (MIT), and meArtificial intelligence (AI), it can predict a file racing of patients through their medical photographs. Ziad Obermayer, Associate Professor at the University of California, Berkeley, Confirm the fact that algorithms see the race It can be dangerous.
The research was conducted using a set of public and private data. Among them Chest x-ray, extremity x-ray, chest CT scan, digital mammogram.
Apart from the above, the team trained a Educational model for identifying the white, black or Asian race Although there is no explicit mention in the patient’s photographs. To achieve this, a series of experiments on artificial intelligence were conducted to investigate the possible mechanisms of race detection.
So, variables like Differences in anatomy, bone density, and imaging resolution, among other things. The study highlighted that despite the variables, the AI continued to detect sweat from chest X-rays.
Leo Anthony Seely, MIT researcher and assistant professor of medicine at Harvard Medical School, He commented that algorithms can amplify existing inequalities. Therefore, he considered it important to think and reconsider whether humanity is ready to bring artificial intelligence to the patient’s bed.
The study, called “Patient Ethnicity Recognition in Medical Imaging Using Artificial Intelligence: A Modeling Study,” was published in Lancet Digital Health May 11, 2022. Seely and Marzeh Gassemi wrote the article with 20 other writers from four countries.
Scientists first showed that Artificial intelligence He was able to anticipate the races in Various photographic methods. Similarly, such a mechanism Guess the different data sets, clinical assignments and academic centers.
For this purpose, three large sets of chest radiographs and The model was tested on an invisible subset of the data used. Next, they trained racial identification models on non-chest X-ray images of multiple sites on the body. This is amazing In order to see if the performance of the model is limited to chest radiographs.
In order to explain the behavior of the model, the team covered several bases. Among them, the differences in physical properties between Different ethnic groups, disease distribution, and site- or tissue-specific differences. Likewise, the effects of social biases, environmental stress were covered in the study, the ability of artificial intelligence to detect races and whether regions of the image contributed to their identification.
In this way, scientists discovered that the ability of artificial intelligence to predict racing As per diagnostic nomenclature was lower. For their part, the Models based on chest X-ray images have a better prediction.
However, scholars have realized that the availability of ethnic identity labels is Limited. As a result of the scenario described above, focus on it Asian, black and white population.
Another work by Qasimi and Seely, directed by the MIT student, Hammad Adamfind out that Artificial intelligence The genders declared by the patient can also be identified. This is from clinical observations, even when these observations lack clear indications of race.
Download the note
Follow us on Google News
“Social media evangelist. Student. Reader. Troublemaker. Typical introvert.”