One in nine women within the developed world might be diagnosed with breast cancer in some unspecified time in the future in her life. The prevalence of breast cancer is increasing, an effect caused partially by the fashionable lifestyle and increased lifespans. Thankfully, treatments have gotten more efficient and more personalized. Nevertheless, what is not increasing – and is in actual fact decreasing – is the variety of pathologists, or the doctors whose specialization is examining body tissues to supply the particular diagnosis obligatory for personalized medicine. A team of researchers on the Technion – Israel Institute of Technology have due to this fact made it their quest to show computers into effective pathologists’ assistants, simplifying and improving the human doctor’s work. Their recent study was recently published in Nature Communications.
The particular task that Dr. Gil Shamai and Amir Livne from the lab of Professor Ron Kimmel from the Henry and Marilyn Taub Faculty of Computer Science on the Technion set out to attain lies throughout the realm of immunotherapy. Immunotherapy has been gaining prominence in recent times as an efficient, sometimes even game-changing, treatment for several forms of cancer. The premise of this type of therapy is encouraging the body’s own immune system to attack the tumor. Nevertheless, such therapy must be personalized as the right medication have to be administered to the patients who stand to profit from it based on the particular characteristics of the tumor.
Multiple natural mechanisms prevent our immune systems from attacking our own bodies. These mechanisms are sometimes exploited by cancer tumors to evade the immune system. One such mechanism is expounded to the PD-L1 protein – some tumors display it, and it acts as a kind of password by erroneously convincing the immune system that the cancer mustn’t be attacked. Specific immunotherapy for PD-L1 can persuade the immune system to disregard this particular password, but in fact would only be effective when the tumor expresses the PD-L1.
It’s a pathologist’s task to find out whether a patient’s tumor expresses PD-L1. Expensive chemical markers are used to stain a biopsy taken from the tumor so as to obtain the reply. The method is non-trivial, time-consuming, and at times inconsistent. Dr. Shamai and his team took a unique approach. Lately, it has turn out to be an FDA-approved practice for biopsies to be scanned so that they might be used for digital pathological evaluation. 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 in fact, we needed to prove them incorrect.”
Neural networks are trained in a way much like how children learn: they’re presented with multiple tagged examples. A baby is shown many dogs and various other things, and from these examples forms an idea of what “dog” is. The neural network Prof. Kimmel’s team developed was presented with digital biopsy images from 3,376 patients that were tagged as either expressing or not expressing PD-L1. After preliminary validation, it was asked to find out whether additional clinical trial biopsy images from 275 patients were positive or negative for PD-L1. It performed higher than expected: for 70% of the patients, it was capable of confidently and appropriately determine the reply. For the remaining 30% of the patients, this system couldn’t find the visual patterns that may enable it to choose by hook or by crook. Interestingly, within the cases where the synthetic intelligence (AI) disagreed with the human pathologist’s determination, a second test proved the AI to be right.
It is a momentous achievement. The variations that the pc found – they are usually not distinguishable to the human eye. Cells arrange themselves in a different way in the event that they present PD-L1 or not, however the differences are so small that even a trained pathologist cannot confidently discover them. Now our neural network can.”
Professor Ron Kimmel, Henry and Marilyn Taub Faculty of Computer Science, Technion-Israel Institute of Technology
This achievement is the work of a team comprised of Dr. Gil Shamai and graduate student Amir Livne, who developed the technology and designed the experiments, Dr. António Polónia from the Institute of Molecular Pathology and Immunology of the University of Porto, Portugal, Professor Edmond Sabo and Dr. Alexandra Cretu from Carmel Medical Center in Haifa, Israel, who’re expert pathologists that conducted the research, and with the support of Professor Gil Bar-Sela, head of oncology and hematology division at Haemek Medical Center in Afula, Israel.
“It’s an incredible opportunity to bring together artificial intelligence and medicine,” Dr. Shamai said. “I like mathematics, I like developing algorithms. With the ability to use my skills to assist people, to advance medicine – it’s greater than I expected after I started off as a pc science student.” He’s now leading a team of 15 researchers, who’re taking this project to the following level.
“We expect AI to turn out to be a robust tool in doctors’ hands,” shared Prof. Kimmel. “AI can assist in making or verifying a diagnosis, it might probably help match the treatment to the person patient, it might probably offer a prognosis. I don’t think it might probably, or should, replace the human doctor. But it might probably make some elements of doctors’ work simpler, faster, and more precise.”
Source:
Technion-Israel Institute of Technology
Journal reference:
Shamai, G., et al. (2022) Deep learning-based image evaluation predicts PD-L1 status from H&E-stained histopathology images in breast cancer. Nature Communications. doi.org/10.1038/s41467-022-34275-9.