Researchers at Mount Sinai Hospital in the United States have announced that they have developed an image recognition AI that can diagnose COVID-19 infection from patient lung CT images.
The AI has its own algorithm to analyze lung CT images with parameters such as patient age and symptoms, blood tests, and contact history with infected people. The researchers say they have great potential to quickly analyze large amounts of data. Mount Sinai Hospital has revealed that it has been able to utilize its expertise to implement CT data of COVID-19 patients with AI through cooperation with Chinese medical institutions.
The research team said that the AI model has the same accuracy as radiation with extensive diagnostic experience, and in some cases, CT images show no clear signs of disease, but in some cases, they have been able to show excellent analysis results.
The research team explains that the AI is an extension of a previous study that looked at the characteristic patterns in the lungs of COVID-19 patients and how they change in a week and a half. Image recognition AI is trained using CT images of more than 900 patients imported from Chinese medical institutions, and patient clinical information is also integrated into the AI algorithm to predict the final decision for the diagnosis of Corona 19.
For example, patients with COVID-19 may have symptoms, but CT does not necessarily show signs, and they may be mistaken for flu and cold with strong symptoms. However, in the case of this AI, the radiologist was able to detect 68% of the patients who tested negative after viewing the CT image. This is also important for recognizing early infections, isolating them and not creating new infections.
In fact, CT scans are not used to diagnose COVID-19 in the United States. However, medical institutions can also avoid wasting resources if they can be quarantined immediately if they have patient CT and are positive if necessary. In addition, AI can be used as a second opinion as it takes time to determine laboratory tests such as PCR tests, and ambiguity such as false positives and false negatives may still appear. However, the research team at Mount Sinai Hospital plans to further develop an AI model to find clues about the selection of effective treatments due to subtle differences between CT denatures and clinical information. Related information can be found here .