According to research published in Nature Medicine, deep learning, a form of artificial intelligence, can detect malignant small tumors in the lungs through low-dose chest CT scans.
A study just published in Nature Medicine shows that deep learning, a form of artificial intelligence (AI), can detect malignant small tumors in the lungs through chest x-ray CT scans. low (LDCT).
Google scientists developed a deep-learning model and applied 2,763 sets of unidentified CT scans, provided by Northwestern Medicine, to test the accuracy of their new system.
They discovered that this AI-powered system could detect some microscopic lung tumors with the AUC model in 0.94 test cases.
These cases were taken from the Northwest Electronic Data Warehouse, as well as other data sources from Northwestern Medicine.
The deep learning system optimizes the most important CT scans and previous CT scans of the patient, taking this as input.
Previous CT scans may be helpful in predicting malignant lung cancer, since the increased incidence of suspicious lung tumors may be an indicator of malignancy.
Computers that are programmed to use LDCT scans have not been fully confirmed for patients who have been confirmed through a biopsy.
The new system also identifies a region of interest and determines whether the area is at high risk for lung cancer.
The model is 6 times more effective than an X-ray machine in cases where previous pictures cannot be diagnosed and equally effective as X-ray machines when the disease is identified.
Co-author of the study, Mozziyar Etemadi, said: "This system can classify a lesion more specifically. We can not only better diagnose cancer patients, but we can also identify corpses. to identify people without cancer, so they do not have to carry out expensive and high-risk lung biopsies. "
However, the researcher warns that this finding should be clinically confirmed in people who already have the disease.
Lung cancer is the main cause of cancer deaths in the US, about 160,000 people in 2018.
Large clinical trial programs across the United States and Europe have shown that chest CT can identify cancer and reduce mortality.
However, the error rate is still high and the ability to access this method makes many cases of lung cancer often detected in the late stages and difficult to handle.