Blog

This is your place to catch up with the characters and conversations that make 3Scan such an inspiring place to work, and work with.

 

Human enabled, machine enhanced
Automation as a framework for computer assisted diagnosis in pathology
By Dr. Navid Farahani

Pathologists have used light microscopes to examine tissue sections on slides since the mid-19th century. Until very recently, this method was the only one available to look at these histological preparations. However, as new technologies and microscopy techniques are developed, pathology is becoming more digital, and a new arsenal of tools is available to pathologists. One of these tools is whole-slide imaging (WSI), wherein conventional slides are scanned to produce digital slides. Once these slides are scanned, specialized software allows pathologists to view the samples in a virtual slide viewer. This has the potential to greatlyspeed up a pathologist’s job. However, an even more exciting technology is tantalizingly close, and would massively reduce the time involved in diagnosis and routine tasks. While still theoretical, the pathology machine-enabled diagnosis (pMED), an intelligent computer assistant, would be able to view whole-slide images and measure things like tumor size, make lymph node counts, and flag regions of interest (ROIs) for a human pathologist to take a closer look at. It could even write most of the pathologist’s report, leaving the pathologist more time to do the important work of diagnosis. While this may seem far-fetched, complex automata already exist—consider Google’s (Mountain View, California) self-driving car project or Amazon’s (Seattle, Washington) inexpensive autonomous flying drone aircraft. In the medical field, a computer-screened Papanicolaou test is already used in cervical cancer screening.

A highly sophisticated pMED would require a combination of advanced image analysis modalities, well-designed human-computer interfaces, and an understanding of the traditional pathology workflow. It would integrate several facets of artificial intelligence technologies, such as computer vision, machine learning, natural language processing, and speech recognition. These technologies would allow pMED to, respectively, identify objects in whole-slide images, learn new information and perform pattern recognition through exposure to data without being explicitly programmed to do so, produce pathology reports in ways humans can read, and respond to commands vocally from a pathologist.

While the technology to have a fully functional pMED is still lacking, Dr. Navid Farahani and colleagues at 3Scan, Inc. in collaboration with Dr. Zheng Liu (Department of Pathology, Barnabas Health, Livingston, NJ) and Dr. Jeffrey Fine (Department of Pathology, University of Pittsburgh Medical Center, Pittsburgh, Pennsylvania), recently developed a prototype of a pMED using slides of a breast cancer specimen. They reimagined the workflow with pMED assisting a pathologist and explained how steps can be automated. This prototype is published as a Powerpoint (Microsoft, Seattle, Washington) presentation, and is publicly available. For more information about this exciting new step in digitizing pathology and to see the prototype yourself, please click on the link to see a full copy of Dr. Farahani’s article.