New I3LUNG project to fund AI tools to improve treatment and outcomes for lung cancer patients

Newswise — Medical University of Chicago has joined the I3LUNG project, a research initiative funded by a five-year, €10 million grant from the European Union to develop a decision support tool to create personalized lung cancer treatment plans. The project will use artificial intelligence (AI) software and machine learning to analyze a wide range of information from clinical data, radiological images and biological characteristics of tumors.

“This highly innovative project aims to use machine learning and artificial intelligence to predict immunotherapy outcomes, which we are currently unable to do,” said Marina Garassino, MD, professor of medicine and head of the University of Chicago’s Medicine Project branch. “At this time, we know that less than 30% of patients will have a good response to immunotherapy, but we are unable to predict who these patients are. Improving our ability to make these predictions can help us better tailor treatments to each patient. »

The data will come from 2,000 patients from multiple research centers in the United States, Italy, Germany, Greece, Spain and Israel. Two hundred new patients will be enrolled in a prospective study to gather new biological data on tumor genetics, immune system, digital pathology, gut microbiome, imaging, and other genetic and molecular analyses. In parallel, the researchers will also conduct a psychological study to integrate patients’ experiences and preferences into decision support tools.

The UChicago Medicine team will apply the expertise of researchers from Garassino’s labs and Alexander Pearson, MD, PhD, a longtime machine learning scientist and clinician, to help process the original data set of patients and prepare it for use by other researchers. The data can in turn be used to develop and test new algorithms to predict immunotherapy outcomes.

“Here at UChicago Medicine, we have a lot of expertise in dealing with these types of data,” said Pearson, assistant professor of medicine. “Finding the most relevant signal in messy genomic data, or finding the growth patterns of a tumor, or the characteristics of a CT scan that correspond to good or bad outcomes when a cancer is treated, all of these areas were led by us, so when Dr. Garassino brought his global expertise in lung cancer to the organization, it was clear that we could work together to take advantage of these emerging methods.

Garassino came to UChicago Medicine during the COVID-19 pandemic, bringing not only her expertise in lung cancer, but also her connections to a multitude of international institutions.

“Our work will use all the patient data we have access to – genetics, histology, lipidomics, proteomics,” Garassino said. “We can use this data to create a huge database of everything that has been collected in the past, and with all this information we can try to build an algorithm and train artificial intelligence to see if we can find a signature that will predict how effective immunotherapy will be for a given patient.

Ultimately, the hope is that the tools developed by the I3LUNG project can be applied not just to treatments for non-small cell lung cancer, but to other types of cancers and treatments.

“When he came on the scene, we had the idea that immunotherapy would be the solution to all cancerous tumors,” Garassino said. “But we now know that’s not true, and really, only some patients respond well. We are very happy with what we have achieved, but we still don’t have a great solution. Tumors are incredibly complex, and with our normal statistical approaches it is impossible to understand this complexity. The promise of AI is that it can learn all of these complex layers and integrate all of this information to help inform individual care.

Work on the I3LUNG project at the University of Chicago will build on expertise in integrating clinical data and AI into translational research, complemented by the work of Pearson, Garassino and others, who are dedicated to using machine learning to improve clinical care.

“Most people in my lab are clinicians first and machine learning experts second,” Pearson said. “This approach means that everything we do is in the service of our patients, rather than just designing algorithms for an algorithm’s sake. It’s really important to get it right for doctors who will be making decisions for their patients using this information. These drugs can have lifelong effects. I hope that with a careful and rigorous approach, we can truly leverage these tools in a safe and valuable way for patient care.