Rami Vanguri says he will “always be a physicist at heart.” And even as a  postdoctoral fellow at DSI, where he’ll explore ways to improve prostate cancer screening, his physics experience will be central to his success.

A physics major at the University of California at San Diego, Vanguri went on for a Ph.D. in elementary particle physics at the University of Pennsylvania. While at UPenn, he had a chance to work at the European Organization for Nuclear Research (CERN), which operates the largest particle physics lab in the world. He participated in the ATLAS experiment at CERN’s Large Hadron Collider, analyzing petabytes of data from particle collisions that occur within the collider at about 600 million times per second. The datasets were so colossal that to analyze them Vanguri used a distributed computing resource called the Worldwide LHC Computing Grid.

Around this time, he began following medical developments, especially how data scientists were applying novel data techniques to improve aspects of healthcare systems. He was intrigued.

“As a physicist trained in computation, imaging, and analytics with large datasets, the idea of applying my skills to a burgeoning new field – big data in medicine – became very tempting to me,” Vanguri says.

For his postdoctoral research project at DSI, he aims to combine data science with magnetic-resonance imaging physics to improve prostate cancer diagnosis and staging. Prostate cancer is the second leading cause of cancer death in American men. His project is ambitious and brings together researchers from across Columbia University: experts in biomedical informatics, biomedical engineering, as well as in clinical departments including radiology, urology, and pathology will all be involved. Working together, they will try to develop diagnostic tools that could help clinicians detect prostate cancer more quickly and less invasively.

Vanguri notes that in order to stage the disease, a patient must be biopsied. He is looking to develop a less-invasive procedure involving computer-vision algorithms on magnetic resonance images. A successful algorithm could assist clinicians in determining the severity and location of the prostate lesions so that the proper treatment or surveillance of the disease can be planned, he says.

He intends to use deep-learning techniques that combine magnetic resonance imaging and data from patients’ electronic health records (EHR) to develop the algorithms, which could make biopsies more efficient and expedite treatment for cases predicted to be severe.

EHRs contain a trove of patient data including medical procedures, health conditions and prescribed drugs, all of which can be correlated to a patient’s prostate cancer pathology, he says. Men with colorectal carcinomas, for example, are predisposed to prostate cancer – and that history would appear on a health record. EHRs also contain data on drugs used to treat Crohn’s disease, which is a risk factor for colorectal carcinomas and may also alter the risk profile for prostate cancers. Additionally, EHRs contain ancestry information correlated to the risk of prostate cancer. African-American men, for instance, are more likely to get prostate cancer – a fact the algorithm will take into account.

“Having valuable EHR data, along with patient imaging data from an MRI, can improve the disease pathology prediction,” he says. “Ideally, if the research works, it would be a big step forward in the diagnosis of prostate cancer.”

Vanguri has three postdoctoral advisers whose specialties illustrate the scope of his project: Nicholas Tatonetti, an expert in big data and electronic health records; Andrew Laine, an expert in  biomedical imaging; and Sachin Jambawalikar, the chief of medical physics who has an expertise in MRI physics and the use of machine learning in a clinical context. He’ll also work with clinicians from Columbia University Irving Medical Center to ensure the scientific and clinical relevance of the study.

“My project crosses disciplines and brings together experts who don’t commonly work together,” Vanguri says. “That’s what so exciting about this research.”

As part of his DSI research, Vanguri also plans to incorporate MRI data and EHRs from some 300 patients into a study curated by Hiram Shaish, Assistant Professor of Radiology at Columbia University Irving Medical Center, and Sven Wenske, Assistant Professor of Urology. He is also working with Renu Virk, Assistant Professor of Pathology & Cell Biology, to include histology slide images from biopsies.

In addition, Vanguri, along with Laine and Jambawalikar, will work with Sairam Geethanath at the Columbia Zuckerman Mind Brain Behavior Institute to promote the development of more advanced MR imaging techniques to detect prostate cancer.

The DSI postdoctoral fellowship is Vanguri’s second at Columbia. In 2015, he accepted a position as a postdoctoral research scientist in the Tatonetti Lab, where he helped develop a database of drug effects, interactions and methods to estimate genetic ancestry using electronic health records. More recently, he was the second author on a paper titled “Disease Heritability Inferred from Familial Relationships Reported in Medical Records,” published in Cell. Now as a DSI postdoc, Vanguri hopes to leverage what he has learned in the Tatonetti Lab, at CERN and from physics to succeed with his prostate cancer project. Advanced algorithms that use deep learning to help analyze medical imaging are changing the way clinicians consider disease pathology, he says.

“With the guidance of my research advisers,” he adds, “I aim to develop algorithms with multi-modal data including MRI and EHRs that can help clinicians detect prostate cancer early, accurately and less invasively with the ultimate goal to extend and improve the quality of patents’ lives.”

 

— Robert Florida