Google has developed a new image analysis technique to help medical personnel in their diagnosis. The first results are already very encouraging, with a rate of discovery that is higher than that of the doctors.
The engineers of Google presented a new use of artificial intelligence in the medical field. The latter, in conjunction with several medical centers, has developed a technique for predicting the risk of a lung cancer in an individual.
An AI compares the studies to precisely detect cancer risks in a patient
The system is based on machine learning. It consists of a database containing thousands of radiographs, and the analysis tool uses them as the comparison point for detects the appearance of tissues and lung nodes in a patient, even almost invisible to the naked eye.
To detect lung cancer, radiologists usually examine hundreds of images of a single CT scan. Google wants them here technical assistance and the first results obtained are optimistic.
Promising Results to Be Confirmed by Other Tests
During the first tests, Google asked his AI to investigate 45,856 screens of thoracic computed tomography. The company then compared the results to those of six radiologists accredited by the representative body of the medical profession.
Finally, the system discovered 5% additional cases compared to human analysis and reduced false positive effects by about 1
Google will continue to improve its device and allow it to pass many other clinical trials before considering offering it to hospitals. The Mountain View Society hopes to increase the patient survey, especially in the US where only 2-4% pass the test during its lifetime.