AI could make doctors' jobs easier, faster and more accurate

AI could make doctors’ jobs easier, faster and more accurate

One in nine women in the developed world will be diagnosed with breast cancer at some point in her life. The prevalence of breast cancer is increasing, an effect caused in part by modern lifestyles and longer lifespans. Fortunately, treatments are becoming more effective and personalized. However, what is not increasing – and in fact decreasing – is the number of pathologists or physicians whose specialty is the examination of body tissues to provide the specific diagnosis needed for personalized medicine. A team of researchers from the Technion – Israel Institute of Technology has therefore set itself the task of transforming computers into efficient assistant pathologists, simplifying and improving the work of the human doctor. Their new study has just been published in Nature Communication.

The specific task that Dr. Gil Shamai and Amir Livne from the laboratory of Professor Ron Kimmel of the Henry and Marilyn Taub School of Computing at the Technion set out to perform falls within the field of immunotherapy. Immunotherapy has gained prominence in recent years as an effective, sometimes even revolutionary, treatment for several types of cancer. The basis of this form of therapy encourages the body’s immune system to attack the tumor. However, such therapy must be individualized as the right drug must be given to patients who will benefit from it based on the specific characteristics of the tumor.

Multiple natural mechanisms prevent our immune system from attacking our own body. These mechanisms are often exploited by cancerous tumors to evade the immune system. One of these mechanisms is linked to the protein PD-L1 – certain tumors display it, and it acts as a kind of password by falsely convincing the immune system that the cancer should not be attacked. Specific immunotherapy for PD-L1 can persuade the immune system to ignore that particular password, but of course would only be effective when the tumor is expressing PD-L1.

It is a pathologist’s job to determine if a patient’s tumor expresses PD-L1. Expensive chemical markers are used to stain a biopsy taken from the tumor to get the answer. The process is non-trivial, time-consuming, and sometimes inconsistent. Dr. Shamai and his team took a different approach. In recent years, it has become an FDA-approved practice to digitize biopsies so that they can be used for digital pathology analysis. Amir Livne, Dr Shamai and Prof Kimmel decided to see if a neural network could use these scans to make the diagnosis without requiring additional processes. “They told us it couldn’t be done,” the team said, “so of course we had to prove them wrong.”

Neural networks are trained in a way similar to how children learn: they are presented with several labeled examples. A child is shown many dogs and various other things, and from these examples he gets an idea of ​​what a “dog” is. The neural network developed by Professor Kimmel’s team was presented with digital biopsy images of 3,376 patients who were marked as either expressing PD-L1 or not. After preliminary validation, it was asked to determine if biopsy images from additional clinical trials of 275 patients were positive or negative for PD-L1. He performed better than expected: for 70% of patients, he was able to determine the answer confidently and correctly. For the remaining 30% of patients, the program could not find the visual patterns that would allow it to decide one way or another. Interestingly, in cases where the artificial intelligence (AI) disagreed with the determination of the human pathologist, a second test proved that the AI ​​was right.

It is a momentous achievement. The variations that the computer found – they are indistinguishable to the human eye. Cells arrange differently whether they have PD-L1 or not, but the differences are so small that even a trained pathologist cannot confidently identify them. Now our neural network can.”

Professor Ron Kimmel, Henry and Marilyn Taub College of Computing, Technion-Israel Institute of Technology

This achievement is the work of a team including Dr. Gil Shamai and graduate student Amir Livne, who developed the technology and designed the experiments, Dr. António Polónia of the Institute of Molecular Pathology and Immunology at ‘University of Porto, Portugal, Professor Edmond Sabo and Dr Alexandra Cretu of Carmel Medical Center in Haifa, Israel, who are expert pathologists who conducted the research, and with the support of Professor Gil Bar-Sela, Head of division of oncology and hematology at Haemek Medical Center in Afula, Israel.

“This is an amazing opportunity to bring artificial intelligence and medicine together,” Dr Shamai said. “I love math, I love developing algorithms. Being able to use my skills to help people, to advance medicine, is more than I expected when I started out as a computer science student. ” He now leads a team of 15 researchers, who are taking this project to the next level.

“We expect AI to become a powerful tool in the hands of doctors,” Prof Kimmel said. “AI can help establish or verify a diagnosis, it can help tailor treatment to the individual patient, it can offer a prognosis. I don’t think it can or should replace the human doctor. But it can provide certain elements of the work of doctors simpler, faster and more precise.”


Technion-Israel Institute of Technology

Journal reference:

Shamai, G., et al. (2022) Deep learning-based image analysis predicts PD-L1 status from H&E-stained histopathology images in breast cancer. Communication Nature.

#doctors #jobs #easier #faster #accurate

Leave a Comment

Your email address will not be published. Required fields are marked *