Researchers at Mount Sinai Hospital in the United States announced that they have developed an image recognition AI that can diagnose the infection of a new coronavirus infection (COVID-19) from CT images of the patient's lungs. This AI has a unique algorithm that analyzes lung CT images along with parameters such as patient age, symptom blood test, history of contact with infected person, researchers said, “ We analyze a large amount of data quickly. It has great potential. " “We have been able to leverage our expertise in implementing CT data for new corona patients in AI in collaboration with Chinese medical institutions,” says Mount Sinai Hospital. The study's lead author, Zahi Fayad, said the AI model was "as accurate as an experienced radiologist, and in some cases CT images showed no clear signs of disease. In this case, we were able to show excellent analysis results. ”
The AI identifies a characteristic pattern in the lungs of COVID-19 patients' lungs and expands on previous studies investigating how it changes after a week and a half, the researchers say. .. While training image recognition AI using CT images of more than 900 patients ordered from medical institutions in China, doctors will also diagnose COVID-19 by integrating clinical information of each patient into AI algorithm Predict the final decision by mimicking a workflow.
For example, a patient with COVID-19 may not be able to find signs in CT even if they have symptoms, and may be mistaken for a flu or a severe cold. However, in the case of this AI, 68% of the patients who were actually positive among the patients who were judged to be negative by looking at the CT image by the radiologist were able to be detected. This is important to recognize that you are infected early and to quarantine it so that no new infected person is created.
In fact, researchers say that CT scans are rarely used to diagnose COVID-19 in the United States. However, it may be possible for medical institutions to avoid wasting resources by taking CT of patients as needed and isolating them as soon as they become positive. In addition, it takes about 2 days to make a judgment in a laboratory test such as a PCR test, and if you still get an ambiguous judgment such as false positive or false negative, you can also provide a "second opinion" by AI. Considering that, this AI seems to be able to be used immediately.
However, researchers at Mount Sinai Hospital intend to further develop this AI model to find clues for effective treatment selection from the subtle differences between CT data and clinical information.
. (tagsToTranslate) Engadget (t) Japan version (t) Engadget (t) PC (t) Mobile (t) Smartphone (t) iPhone (t) Mac (t) Latest (t) information