Proscia is Teaching Computers to Beat Cancer
For 150 years, pathologists have used glass slides and microscopes to analyze tissue samples, understand cancer and determine the best course of treatments. One Baltimore startup, however, is ushering this process and the whole field of pathology into an era of computational pathology.
Founded in 2014 by Johns Hopkins undergraduates from the Whiting School of Engineering, Proscia began developing data management applications that enable researchers and pathologists to upload, extract and manipulate data, and telepathology solutions that allow a large number of people to analyze the same slide at the same time from different locations around the world.
“Pathologists analyze billions of slides each year, and almost every single one of those is analyzed under a microscope,” says Proscia CEO David West. “In the next couple of years, a huge chunk of those images will be digitized before they’re looked at by a pathologist. This represents one of the biggest trends in big data in medicine ever.”
When West says big data, he isn’t exaggerating. Despite the tissue samples fitting onto a 1-inch-by-3-inch slide, their digital high-resolution images are about a gigabyte each. The large files allow researchers and pathologists to enlarge the images to 20 to 40 times their original size and clearly examine components of individual cells.
“It’s such an ungodly amount of information, and in every one of those images is a little piece of the puzzle of cancer,” West says. “What we want to do is start unlocking that.”
A few months ago, Johns Hopkins signed on as a Proscia client, enabling the institution’s pathologists and other researchers to access and tap into pathology data from across the institution. Proscia’s technology also powers Johns Hopkins’ Surgical Pathology Unknowns database. Each week pathology residents receive six to eight unknown surgical pathology cases. These cases typically comprise both classical and unusual lesions seen in surgical pathology the previous week, providing a tremendous learning opportunity.
Though Proscia initially focused product development on the research market, it has increasingly invested in developing image analysis solutions that would benefit pathologists in a clinical setting. The technologies in this sector use computer vision and machine learning to enable pathologists to make more informed and more accurate decisions about the severity of a patient’s cancer and the best course of treatment.
Though Proscia has invested in this area for about a year, the market for these technologies became more attractive in April when, for the first time, the Food and Drug Administration (FDA) approved the first Whole Slide Imaging system for primary diagnostic clinical use in pathology. According to the FDA, this system facilitates the “review and interpretation of digital surgical pathology slides from biopsied tissue.”
“The FDA approval lays the groundwork for the transition from pathology to digital pathology to computational pathology,” says West, who graduated from The Johns Hopkins University in 2016 with a degree in biomedical engineering and a focus on computational biology. “It makes it a good time to be in this space.”
The promise of computational pathology
Science has little room for subjectivity, but unfortunately it still exists. West says two well-educated, experienced and successful pathologists analyzing the same tissue sample could come to two vastly different conclusions. One pathologist may determine that a tumor is likely to metastasize (spread) and want to act aggressively, while the other may not see as grave of a threat. Though pathologists do a terrific job with the information they have, West says machine learning will help them be even better.
“If a human can do a task with enough training, that means a machine can do it with enough training and perhaps more efficiently than a person,” West says. “We basically are training algorithms to look for patterns the same way a human pathologist looks for patterns.”
Proscia’s algorithms provide a quantitative view of cancer. For example, the algorithm could determine that a tumor is 51.5 percent likely to metastasize. Not only does this arm a pathologist with more data when determining treatment, it provides a standardized diagnostic procedure that reduces errors and improves patient outcomes. It also could create significant financial savings for health care providers.
Despite the promise of this technology, West says that image analysis solutions will never wholly replace a pathologist. Instead, he sees the technology as a tool that will automate certain tasks and augment a pathologist’s expertise.
“Studies have shown that combining an algorithm with human knowledge produces the best results,” West says. “You get nearly 100 percent accuracy when you do that.”
Finding success in Baltimore
Proscia has received a lot of recognition for and investments in its work. A little over a year after announcing in February 2016 that it had raised $1 million, Proscia announced another raise for $925,000.Then in June, the Maryland Technology Development Corporation (TEDCO) awarded Proscia the Incubator Company of the Year award for Best Unaffiliated Company.
These successes followed Proscia leveraging multiple resources at Johns Hopkins Technology Ventures (JHTV). West, who entered Johns Hopkins hoping to build his own company, recalled how the support the university provided student startups became much more robust as he worked his way from underclassman to upperclassman and how that support contributed to Proscia’s growth.
“We went through the Ralph S. O’Connor Undergraduate Entrepreneurship Fund, and that was a great experience for us,” West says of the program that supports teams of undergraduate entrepreneurs with up to $10,000 in funding and mentorship. “JHTV is a community. It’s awesome to be a part of it.”
After graduating, West decided to stay in Baltimore because of the resources available to help him build his company. Proscia has operated out of the Spark co-working space near the Inner Harbor for the past year, but West has continued to work closely with JHTV in that time.
“I can say confidently that the support for entrepreneurs and technologists coming out of Hopkins is there,” West says. “In addition to the O’Connor Fund, we made some connections to some funds that were here. I look forward to engaging JHTV more in the future.”