Artificial intelligence in the service of cancer medicine

A technology based on artificial intelligence may quickly and simply decode genetic and molecular information about cancer tumors. On the horizon: personalized medicine for cancer patients

The field of artificial intelligence is gaining momentum in the world of medicine, and also in cancer treatment. At the core of artificial intelligence is deep learning (Deep Learning) - a field of research in computer science whose goal is to create a computer simulation of the operation of the human brain. These are algorithms that allow computer systems to learn from previous examples and experiences - and thus perform a variety of computational tasks.

Prof. Ron Kimmel and Dr. Gil Shamai from the Faculty of Computer Science at the Technion are engaged in geometric processing and learning of signals and images. This is how they train computers to decode visual information on their own. The goals: development of ground, aerial, atmospheric and medical instruments. Dr. Yaki is involved in some of their research Cohen, Director of the Department of Otorhinolaryngology and Head and Neck Surgery at the Rambam Medical Center, and clinical lecturer senior in the Faculty of Medicine at the Technion. In their latest research, which won a research grant from the National Science Foundation, the three collaborated with Ron Slusberg (doctoral student in Prof. Kimmel's laboratory), Dr. Irit Doak (specialist in otolaryngology), Dr. Yoav Binenbaum (specialist in oncology children) and Prof. Ziv Gil (former head of the ear, nose and throat department at the Rambam Medical Center). Together, the scientists and doctors developed a technology based on artificial intelligence, which was nicknamed "computerized pathologist". This technology is used to decode samples (tissues) of cancerous tumors from a genetic and molecular point of view. The system examines the samples and discovers which genetic mutations they contain (which help the development of tumors) and identifies additional characteristic markers (such as proteins). That is, through it you can find out what the unique signature of each tumor is, and thus help decide on the most suitable oncology treatment for the patient, which is based on that signature.

https://youtube.com/watch?v=vZL7LfX4U0A%3Ff

As part of the cancer diagnosis, tissue is taken from the suspicious lump and sent for biopsy in a pathology laboratory. The tissue is examined by pathologists under a microscope and stained with hematoxylin-eosin (H&E) staining - a quick and cheap method designed to distinguish between cells. By looking at the dyed tissue, the pathologists can understand if it is cancerous, what type of cancer it is, how aggressive it is, and more. However, they cannot characterize the tumor in this way from a genetic and molecular point of view - information that is essential for doctors to decide on the most appropriate type of treatment. That's why there are more advanced technologies, such as NGS (which decodes the DNA sequences in the tumor and locates the mutations in it). These are expensive technologies that require additional personnel skilled in performing them, and they do not yet exist in all underdeveloped countries.

Using the technology they developed, the scientists and doctors were able to show for the first time that artificial intelligence is able to extract genetic and molecular information from an analysis of the tissue structure as it is reflected in H&E images. Says Prof. Kimmel, "The computerized pathologist can characterize cancer cells genetically and molecularly according to the shape of the tumor tissue and its environment. That is, based on the analysis of its morphology. Unlike the human eye, however skilled it may be, artificial intelligence is able to identify this unique signature of the tumor and draw conclusions from it ".

Using the technology they developed, the scientists and doctors were able to show for the first time that artificial intelligence is able to extract genetic and molecular information from an analysis of the tissue structure

The technology the researchers built was "trained" on 20,000 H&E images taken from approximately 5,000 breast cancer patients. In about half of them, she was able to determine that there is a molecular expression for the receptors of the sex hormone estrogen, which contribute to the division and spread of the cancer cells, based on an analysis of the shape of the tissue. These receptors are present in most breast cancer tumors and are targeted for specific treatment, one that blocks them or causes them not to receive the hormone they need.

A scan of tissue taken from a breast tumor

"Based on many experiences and examples of images, the computer found certain characteristics in the cancerous tissue - for example, the shape of the cells and the way they arranged themselves - and learned the relationship between them and the molecular and genetic profile of the tumor. In other words, in a simple way, using artificial intelligence, it discovered a unique signature of breast tumors , which is in the form of their tissue. We believe that it will be possible to improve the diagnosis of cancerous tumors and the suitability for oncological treatments, and save complex tests and expensive for detecting genetic mutations and molecular markers," Dr. Shamai points out. Prof. Kimmel adds that "this technology may significantly shorten the diagnosis time and make it cheaper, allow pathologists to focus on more complex work, and promote personalized cancer treatment."

The areas from which the "computerized pathologist" extracted genetic and molecular information about the tumor (colored in red and green)

Today, the researchers are testing the technology on additional cancer tumors, and hope to be able to use it to predict the effectiveness of oncological treatments and chances of survival. "We believe this will be a clinical tool that will help doctors make correct treatment decisions," Prof. Kimmel concludes.

Life itself:

Prof. Ron Kimmel - 57 years old, married and father of four ("ages 14 to 26") and lives in Haifa. He used to be a dancer and now he likes to ride mountain bikes, ski and paint. Among other things, he draws on the iPad and also on walls ("With my daughter. She is the real artist").eature=oembed

Dr. Gil Shammai

Dr. Gil Shamai - 36 years old, lives in Tel Aviv, a graduate of the "Technion program for excellent students". Already as a child he loved mathematics and geometry ("I participated in many competitions in these fields"). His doctorate, which he recently completed in Prof. Kimmel's laboratory, deals with image processing XNUMXD, computational geometry and artificial intelligence in the medical field. Besides, he likes to ride a bike and surf ("Every weekend I ride or Golash"), plays the piano and drums, and loves Iraqi food.

Dr. Yikki Cohen - 56 years old, married and the father of two daughters, lives alternately in Haifa and Tel Aviv. He has been involved in applied research and collaborations with industry for many years. The projects in which he was a partner in the development include XNUMXD endoscopy, a cold plasma device for cancer treatment, and the treatment of voice disorders with the help of Artificial intelligence In his spare time he does boxing and martial arts.

For the article on the National Science Foundation website

More of the topic in Hayadan:

Leave a Reply

Email will not be published. Required fields are marked *

This site uses Akismat to prevent spam messages. Click here to learn how your response data is processed.