July 25, 2021

This is how artificial intelligence has gone from diagnosing cancer to treating it. Science

This is how artificial intelligence has gone from diagnosing cancer to treating it. Science

The new labor revolution, led by the big data and by the artificial intelligence programs, it is different from the previous ones. Doctors, lawyers, publicists, managers … workers who historically have been able to follow the automation with mere curiosity are already tested by their ability to collaborate with the machines. In medicine, digital intrusion arrived first in the field of diagnosis, with algorithms capable of detecting diseases such as Alzheimer's or skin cancer more reliably than the doctors themselves. Now, artificial intelligence graduates from diagnosing diseases to treating them.

"It is relatively easy to validate screening programs: if the computer predicts that a person has skin cancer and you prove that they have it, the model works," explains Aaron Babier, a researcher at the University of Toronto who has created A program able to design personalized cancer treatments. "With treatment algorithms it is more ambiguous: ours worked very well in terms of statistics, but a doctor may see the doses of a treatment plan [generado por el ordenador] and thinks that they do not have the optimal distribution. It's more subjective, "he adds.

The advantages of a treatment generated by artificial intelligence can be considerable. Faced with the hours or days that usually lead a medical physics specialist to devise a radiotherapy plan for a patient with cancer, a computer can do it in minutes, previously trained with images and results of historical cases. Also, if the program evaluates the tumor response after each session -as it does one designed by researchers at the MIT Media Lab-, you can self-regulate your future recommendations to keep doses to a minimum. Without compromising the expected results, this process reduces the toxicity and side effects of the treatment. Other computer program he created complete radiotherapy plans that turned out to be better than those recommended by specialists in 83% of the cases, according to the blind evaluation of two oncologists recruited by the creators of the model at the University of California.

Much of this innovation occurs abroad, led by private companies such as IBM, a pioneer in artificial intelligence applied to oncology thanks to its star robot Watson, and Google, which collaborates with public hospitals of University College London in the United Kingdom to test their algorithm generator of personalized radiotherapies, DeepMind Health. But technology is still in its infancy. The Watson supercomputer sometimes recommended cancer treatments "Dangerous and incorrect", according to the US health magazine STAT, which had access to internal IBM documents. Olaf Ronneberger, a senior researcher at the DeepMind Health team in London, says the results of the Google program are "very promising", but artificial intelligence is still far away from cancer patients.

The use of artificial intelligence to treat cancer for now is limited to pilots

Companies like the Spanish Quibim they already work with advanced medical imaging algorithms to detect changes produced by the disease or treatment. In the clinic, these applications are confined to diagnosis and monitoring; The image analysis carried out by the machines helps the oncologists to optimize their treatment, but in no case they substitute them. "The machine delimits the lesion's pixels much better than the human eye, but we still have to verify that the software is not wrong," explains Francisca Mulero, head of the molecular imaging unit of the National Center for Oncological Research (CNIO) The impact that artificial intelligence will have on the sector is difficult to foresee, given that for now it is limited to pilots far from the clinical routine. However, rapid technological advances bring to light the problem of digital mismatch, the asymmetry between the technological skills required and the real skills of professionals.

A recent study by the European Commission shows that nine out of ten jobs will require digital skills in the next decade, but 44% of Europeans between the ages of 16 and 74 do not have the necessary skills to face this transition. However, Babier says that the treatment plans designed by algorithms do not require technological training of oncologists so that they can apply them to their patients: "The additional training needed by doctors is essentially nil. They receive a solution with which they are already accustomed to work, and which is also specific to the situation of their patient. "

Those who do not handle this technology will be the ones who will not work

Mulero agrees with this assessment. "It's like using the mobile. There are applications to manage the bank and the grandmothers use them, although the one who understands it will take better advantage, logically, "he says. For her, the question is one of acceptance among professionals, rather than training, although the second should improve the first. "For example, radiology is one of the fields that has most adapted to the latest technologies. Radiologists will not be replaced, but those who do not use this technology will be the ones who will not work, "says the doctor.

Test: Test your level of 'frikismo'

The Commission has recently produced the Digital Europe Program, a package of measures of nine billion euros to promote areas such as artificial intelligence and, above all, skills and adoption of digital technologies in society. Private companies are involved in this front in Europe and also multinationals such as Facebook or Microsoft are investing to support the competences in different countries of the Union such as Spain, Italy and Poland, which suffer from digital mismatch more sharply, but also in the United Kingdom. However, regardless of the technical knowledge that physicians must acquire, a key to integrating artificial intelligence in oncology lies in the consolidation of multidisciplinary teams. "Many colleagues are training mathematicians, physicists, chemists, engineers. The doctors have to be with them to get more performance, "says Mulero. "When a doctor and a basic researcher come together, they are unstoppable."


Source link