Abstract
New startups are developing pattern-recognition algorithms that could one day help pathologists more accurately spot tumors on digitized tissue images, thereby aiding in diagnosis, treatment, drug discovery, and more.
An early pioneer of computational pathology, Thomas Fuchs, PhD, of Memorial Sloan Kettering Cancer Center (MSKCC) and Weill Cornell Medicine, both in New York, NY, announced the launch of a new academic spin-off company in February. The start-up, known as Paige.AI, aims to replace pathology's glass slides and microscopes with digitized images and artificial intelligence (AI).
It's hardly alone in this endeavor. Software giants, medical device manufacturers, and dozens of small start-ups are developing pattern-recognition algorithms to help pathologists more accurately spot tumors on digitized images of tissue. However, thanks to $25 million from investors, intellectual property from MSKCC, and exclusive rights to the cancer center's library of 25 million pathology slides—one of the world's largest tumor pathology archives—Paige.AI has established itself as a front-runner in the development of a clinical-grade AI tool to guide cancer diagnosis and treatment.
“We have data at a scale no one else has,” says Fuchs, the company's founder and chief scientific officer. Paige.AI is first deploying its machine-learning models, image classification software, and high-performance computers to diagnose prostate cancer. Algorithms for breast cancer, lung cancer, and other tumor types are also in development.
Fuchs sees biomarker analysis as one application of the technology: “The goal is to come up with new grading schemes and scores that can better stratify patients for different therapies.” Additionally, he hopes AI can “help the pathologist be faster, more accurate, and more reproducible.”
Another computational pathology start-up, PathAI, founded by Andy Beck, MD, PhD, recently won an international challenge that pitted algorithms from 23 teams against each other. Beck's algorithm—developed when he was on the faculty at Beth Israel Deaconess Medical Center in Boston, MA, with his lab and a colleague from the Massachusetts Institute of Technology in Cambridge—outperformed expert pathologists when it came to detecting metastases in lymph node slides from women with breast cancer (JAMA 2017;318:2199–210).
Now backed by $15 million in venture capital, Beck and his PathAI team aim to turn their expertise in pathology, biomedical informatics, and computational image analysis into commercially viable software that informs clinical decision-making, aids in drug development, and enables low-cost diagnostics in the developing world.
A number of obstacles stand in the way of the adoption of AI technology in clinical practice. First, to use the algorithms, slide digitization will need to become part of the primary diagnostic workflow, not an afterthought for archival purposes, says Jeroen van der Laak, PhD, of the Radboud University Medical Center in Nijmegen, the Netherlands.
In addition, with few applicable billing codes for AI services, “how do you deploy this in the clinical pathology workflow in a way that allows you to make money?” asks Anant Madabhushi, PhD, of Case Western Reserve University in Cleveland, OH. “That's not clear to me.”
Beck acknowledges these challenges need to be overcome for computer-assisted diagnostic systems to be embraced. However, immediate opportunities for AI-powered pathology abound: For example, Bristol-Myers Squibb, one of PathAI's customers, is using the company's machine-learning algorithms to analyze tissue samples for a better understanding of drug responses in clinical trials.
“There's an explosion of data we can now generate from patient samples,” Beck says. “We can bring huge value to drug development today.” –Elie Dolgin
For more news on cancer research, visit Cancer Discovery online at http://cancerdiscovery.aacrjournals.org/content/early/by/section.
RSS Feeds