This is how an AI engineer imagines the profession of radiologist in the future

This is how an AI engineer imagines the profession of radiologist in the future

Artificial intelligence (AI) has already shown encouraging results in the health sector. And in particular to anticipate or detect non-visible tumors using traditional medical imaging methods. Recently, researchers at the Massachusetts Institute of Technology (MIT) in the United States have developed an AI capable of detecting breast cancer cases four years before they are visible in traditional imaging, according to a study published in the scientific journal Radiology.

Will AI replace or radically change certain professions in the medical profession such as radiologists? On the sidelines of the “France is AI” conference organized on Wednesday, October 23, 2019 at Station F in Paris, Maria Vakalopoulou, a researcher specializing in AI at CentraleSupélec, shared her predictions on the issue with Business Insider France. Here’s what she replied: AI is not supposed to “replace people, but rather help them. What could be really good for a radiologist is in particular to spend less time doing things that can be done by an AI or to have a second opinion thanks to technology.”

The trained engineer believed that AI technologies “can be easily integrated into the daily work of radiologists. And again, it is especially to help them because in the end, the doctors will make the final decision about the treatment to be followed and the disease. I don’t think radiologists will be replaced by AIS, because human lives are at stake. But they will have to be trained in these new tools and be able to interpret them correctly.”

Also read – We tell you how Elon Musk thinks putting chips in people’s brains

Maria Vakalopoulou said doctors “should have training on AI but without necessarily knowing all the algorithmic details. They will need to know all the possibilities that AI offers, be able to work with them and interpret them. And they already do, they have programs that they use to help them, to view images for example. AI is just an improvement that they can have in their toolbox.”

The obstacles holding back the full adoption of AI

However, AI technologies currently present more or less significant obstacles, which hinder their full adoption in hospitals in particular. Among them, Maria Vakalopoulou cited :

  • the limited number of data samples to train AI on some rare diseases ;
  • the lack of annotations of doctors on the pictures : “even if I have a picture, I cannot determine where the tumor is for example, I need the opinion of a doctor, who will then create these annotations. That’s why we always work in collaboration with doctors,” explained the engineer ;
  • annotations may vary depending on the doctors who wrote them ;
  • images that have different appearances because taken with different sensors, under different parameters : “The data we have can be very heterogeneous, due to the different modalities of images that show things differently. In medical imaging, all these modalities (CT-Scan for example) are useful to establish a complete diagnosis. This is a limitation, because you cannot use an AI model that you have trained on one modality and apply it to another. We need additional processes for this,” said Maria Vakalopoulou ;
  • the AI works well but it is sometimes difficult to decipher how it works, which is akin to a black box for doctors.

In any case, AI technologies seem to have a bright future ahead of them in the health sector. Indeed, according to a report by Business Insider Intelligence, spending on AI in healthcare is expected to grow by 48% per year between 2017 and 2023.

Receive our latest news

Every day, the main Business Insider news

Go to our cases Get a free quote