Three young entrepreneurs at Proscia believe AI has the potential to change how pathology is being practiced and ultimately, how cancer is diagnosed
“I was fortunate to work with some well-known researchers during my undergraduate years, demonstrating how algorithms could be used to recognize complex patterns in images of tissues and predict cancer outcomes,” says David West, co-founder and CEO of Proscia, a digital and computational pathology company headquartered in Philadelphia. “However, back in the early 2010s, it was challenging to capture the tremendous amount of data required to train deep learning models. To put it bluntly, the infrastructure to deploy AI models was non-existent. Storage was expensive and image management was cumbersome. More importantly, the regulatory environment for digital pathology was not mature enough to support related activities.That was when we saw an incredible opportunity to build a great business.”
West duly founded Proscia with his classmate Nathan Buchbinder and childhood friend Coleman Stavish in July 2014, after he graduated from John Hopkins University where he studied Biomedical Engineering. The trio built the foundational infrastructure that laboratories would need in order to go digital and a pipeline of algorithms that would eventually become Concentriq, Proscia’s signature digital pathology platform.
“Concentriq enables digital images of biopsies – huge digital files that have historically been hard to manage – to be stored, examined, annotated, diagnosed and shared with other pathologists automatically,” explains West. “It also provides analytics on the process of evaluating digital pathology images.”
The platform was launched in May 2015 as a free solution and soon caught the attention of fellow pathologists. Around the same time, the trio also garnered a $2 million seed funding to pursue Proscia full time to develop an enterprise version of the software and set themselves a long-term goal.
“We have a mission to perfect cancer diagnosis with intelligent software that changes the way pathology is being practiced,” says West. “For the past 150 years, a cancer diagnosis means subjective interpretation of what the human eye could find under a microscope. We want to leverage AI and machine learning powered computational tools to locate information that’s hidden in every tissue sample.”
However, pathologists continue to use conventional slides and microscopes for cancer diagnosis for two reasons. Firstly, the optics of a microscope provides a superior user experience because pathologists can flexibly move the slide and there are pre-set zoom levels. Whereas using a computer is like using a Google map – one must constantly zoom in and out, using a keyboard or mouse to navigate around clumsily. Secondly, digital examination of images requires all physical slides to be digitized. The radiographic films need to be developed and then scanned in on a flatbed scanner with the result that the examination process becomes slow and inefficient.
It means the pressure on laboratories is incredibly intense with increased turnaround time for lab results. In addition, the pathologist workforce is declining by 12% every decade, while biopsy volume is increasing. It’s not hard to see why West describes most pathologists as spending time looking for that needle-in-a-haystack instance of tumor in an otherwise benign case. As such, Proscia’s first AI product DermAI aims to accelerate the work of pathologists as they are now able to look at fewer images and spend more time communicating with patients and their care teams (oncologists or surgeons).
“The solution can be plugged into the Concentriq platform to classify images from dermatology biopsies as ‘likely or not to be cancerous’; single out a potentially cancerous image by giving it a score or find the most likely image among several for a pathologist to look at or draw a circle around the suspicious part of the cell in an image,” West explains. “An absence of common data standard for pathology images (unlike radiology which uses the DICOM standard for decades) is challenging digital pathology. Thus, we are spending more of our time dealing with these challenges that are hindering the delivery of AI In actual laboratories, beyond what’s postulated in interesting academic papers.”
The three founders were named in Forbes’ annual 30-under-30 people in healthcare for 2019. Last December, Proscia landed $23 million in Series B funding led by Scale Venture Partners with participation from Hitachi Ventures. West’s team are planning to use the money to build out its present commercial effort and global sales and marketing. They also planto seek FDA and CE mark clearances.
“My generation is the first that grew up fully connected,” West continues. “It may be a bit of a generalization, but healthcare software, especially in the digital pathology space, is often well below the enterprise design standards we see in every other industry. Because healthcare is a complicated beast, so it can’t be just about looking pretty, but about being powerful and consistently working well. Pathology runs on productivity, so there’s no tolerance for clunky experiences.
“There is no doubt, AI is probably one of the most powerful tools we’ve ever created. It has already opened huge opportunities. Ten years from now, digital pathology will just be pathology. The standard of care will shift; it will become more fluid and democratized for the patients since physical constraints are eliminated. I also expect computer systems to address the bulk of straightforward cases in certain subspecialties. This will reduce the workload of the already decreasing population of strained pathologists – a challenge that, in some countries, we are already witnessing right now.”