Determining the type of cancer of a patient is somewhat complicated in some cases and a reason why pathologists. Oncologists do not always agree, putting the patient’s health at risk if a mistake is made. Fortunately, Google may resolve these issues in the future. Give a diagnosis with 89% of reliability, which contrasts with 73% of a doctor, a figure that falls to 48% when it comes to a diagnosis among several experts according to the documents that the company handles.
Artificial intelligence has very advanced image recognition software. The same that Google has used for its autonomous cars, only in this case it is used to examine. The high-resolution images of cells taken from patients to check if they are affected by cancer. It is a process that pathologists and oncologists also carry out when making a diagnosis, although they cannot count on a database of billions of images to take as a reference to determine which variant of the disease is affecting the patient.
The machine then learns to diagnose by comparison they are given a multitude of referents which they can understand learn. Then be able to dump their knowledge when receiving a new image of a possible case of cancer. Of course this system cannot replace experts in the disease however exact it may be as there is still a large percentage of false positives.
Moreover the strangest variants of the disease are not detected something that an expert in she can see with the naked eye. Its aim then is to serve as an initial analysis so that once given the opinion of the machine the expert can save time by checking what he says or examining points for analysis to corroborate or contradict the diagnosis.