Google’s AI System in Detecting Breast Cancer

Incorporating AI into medicine has been an elusive dream until the present. The failure of systems like IBM Watson in its try at providing healthcare has done more to set AI back in the field than anything else. However, recent research incorporating Deep Mind AI with mammogram imaging technology has shown promising results. Deep Mind AI joined with Google Health in September 2019, and the experiment charged the AI with spotting breast cancer regions from mammograms. The AI succeeded in determining the incidence of breast cancer in a set of test images with the same accuracy as professional radiologists.

Removing Human Error

Radiologists tend to miss about one-fifth of the early detection of breast cancer images and have been known to give false positives with surprising regularity. By incorporating AI into the mammograms, radiologists may have an easier time determining if a structure should be declared as a sign of breast cancer or not. With breast cancer affecting one in every eight women in the world at this point, higher accuracy in detection is crucial to offer victims the best chances at survival.

The researches trained the system using mammograms from both the US and the UK. A total of more than five thousand images were used in developing the AI’s proficiency. The AI picked up the incidence of breast cancer with surprising accuracy and even reduced the rate of false positives by 5,7% in American patients and 1.2% in those from the UK. Additionally, the system reduced the generation of false negatives – where a patient is told they have no cancer when a tumor does indeed exist. In a separate test, the AI performed admirably when pitted against six other radiologists.

Still Some Issues to Work Out

While the system is promising, it is essential to remember that these studies form part of the first push to incorporate AI into cancer detection. There is no indication that the system will increase the level of patient care. More testing will need to be done, especially in more challenging cases, to determine if the AI has what it takes to operate in the real world.