An investigation published in Lancet DigitalHealth ensures that an artificial intelligence would help detect breast cancer with a higher success rate to that achieved by the medical staff alone. Although many may take it as an “attack” on the wisdom of these professionals, nothing is further from the truth. This AI has been created with the intention of saving lives.
The artificial intelligence used in the tests was developed by Vara. The latter is an emerging company, and everything seems to indicate that it will be very successful in the near future. According to statistics, Vara’s AI has already been used in a quarter of detection centers breast cancer in Germany; country where the company also originated.
How does Vara’s artificial intelligence work? Since MIT Technology Reviewcomment that AI currently has two modes. With the first one, she is able to analyze the X-rays herself. In her second, she helps to differentiate which are those results that appear to be normal, and which could reflect a problem for the patient’s health.
Of course, for the final decision, AI sends its results to radiology professionals. In this way, the sample can be examined more thoroughly, thus avoiding a false positive by artificial intelligence. In the event that the samples sent contain traces of cancer that could have been overlooked by the doctor, the AI will then send an alert for a new review.
This is how Vara’s AI works
To train the AI, Vara has used samples from more than 367,000 mammograms. In addition to the images, these samples also include information about the patient’s history with the disease, the initial diagnoses and the notes of radiologists in charge of reviewing them. Thus, they have achieved that this artificial intelligence can separate the samples obtained into three different categories: “safe cancer”, “unsafe” and “normal safe”.
The results thrown by the AI were then compared to 82,851 other mammograms from different centers. The latter, in addition, were not part of the first study, so intelligence did not know the data offered in advance.
Vara’s results in the real world
Radiologists who check mammograms are known to fail to detect 1 in 8 cancer diagnoses. Something totally normal given that we are talking about human beings. Aspects such as fatigue, work overload and other personal problems can affect the result of the diagnosis.
However, with the help of Vara’s artificial intelligence, the radiologists yielded an improvement in detection of 2.6%. This applied only to radiographs related to breast cancer, of course. The AI was able to detect details that were usually overlooked by radiologists under any of the above situations.
The future of medicine is drawing near
Currently, we know that AI is not capable of detecting cancer by itself. In fact, when he tried to do it, he got poorer results than when the radiologists did it on his own. Of course, he is capable of detecting signals that could be translated into cancerous tissue in the breast areas.
But this is not all. Detecting those that show no signs of cancer, Helps greatly reduce workloads of the medical professional. A very appreciated help when you have to review hundreds of sheets a day, every day of the week.
Do doctors trust the results that Vara calls “normal”? It looks like it is. According to Thilo Töllner, a breast cancer radiologist at a center in Germany, “normal is almost always normal.” In this way, he admits that he spends more time dissecting the positive detections of intelligence, than those that he has described as “normal”.
Although AI is currently unable to work alone, it is expected that in the future this technology will advance enough to do so. Curtis Langlotz, director of the Stanford Center for Artificial Intelligence in Medicine and Imaging, believes that we are facing a totally innovative technology.
This study doesn’t change that, but in the proposed AI-driven process nearly three-quarters of screening studies did not need to be reviewed by a radiologist, while improving accuracy overall.
Curtis Langlotz
Today, AI is already being used in specialized cancer detection centers in Germany. Likewise, this year saw his arrival at a major hospital in Mexicoand another in Greece.
Of course, the plan is to bring this revolutionary intelligence to every corner of the world, and we love seeing the shape these technologies are taking and how much they can bring to humanity. Instead of looking at them with fear and prejudice, we should first seize them for what they are: tools with capabilities we’ve never known.