U.S. Patent 10,496,742 addresses techniques for streamlining pathologist review of whole slide images to support artificial intelligence development
Proscia, a leading provider of AI-enabled digital pathology software, has been granted U.S. Patent 10,496,742 by the United States Patent and Treasury Office (USPTO). The patent relates to techniques for evaluating medical images, including whole slide images, in a networked computer environment to streamline pathologist review and accelerate the collection of data to train and validate artificial intelligence (AI) applications.
For more than 150 years, the standard of care for diagnosing biopsies has relied on a pathologist’s interpretation of patterns in tissue as viewed under a microscope. The limitations of this manual and subjective practice have given rise to digital pathology in which specialized scanners capture whole slide images of tissue samples prepared on glass slides. Whole slide images can be efficiently viewed, managed, and shared on Proscia’s Concentriq® digital pathology platform. They can also be analyzed by computational solutions leveraging artificial intelligence, including Proscia’s pipeline of AI applications, to improve accuracy, drive efficiency, and generate new insights into how we understand cancer.
The successful development of AI depends on ground truth data for training and validating an algorithm. Collecting this data is especially challenging in pathology. An expert pathologist is required to provide a diagnosis, which is inherently subjective. To overcome this subjectivity, it is critical to achieve consensus among pathologists; however, consensus is often difficult to capture given that the current manual slide review process requires significant human intervention.
U.S. Patent No. 10,496,742 covers techniques for a computer-implemented system of evaluating whole slide images to streamline pathologist review and support the collection of training and validation data for artificial intelligence development. It describes techniques that allow for whole slide images to be efficiently shared through a digital review platform as well as addresses the creation of templated forms, which pathologists can use to capture diagnostic data and regions of interest. Systems as disclosed by the patent also include a centralized server for collecting this data from pathologists and storing it for use. Such systems can use collected data to train an artificial intelligence algorithm whereby they receive a new whole slide image and apply machine learning trained on the data to output a diagnosis.
This patent reflects Proscia’s commitment to changing the way the world practices pathology with intelligent software. The Concentriq digital pathology platform incorporates techniques covered by the patent to capture diverse, high quality data for training its pipeline of disease-specific AI applications. Pharmaceutical companies, academic medical centers, and clinical laboratories can also leverage Concentriq to collect data for training and validating their own homegrown AI algorithms.
U.S. Patent No. 10,496,742 is the third granted to Proscia in the past year. The company was granted U.S. Patent No. 10,460,150, which relates to AI-powered dermatopathology, in October 2019. In September 2019, Proscia was issued U.S. Patent No. 10,346,980 covering techniques for processing and analyzing medical images critical to the advancement of digital pathology.