Portfolio Assignment – News and Timeline

Integration of Computational Clinical Support Systems in Pathology

News: Paige Announces World’s First Clinical-Grade Artificial Intelligence in Pathology in Businesswire

With the FDA approval of whole slide scanners in 2017, the field of pathology had experienced a shift in the existing workflow as institutes began digitizing histological scans of tissues– such as cancer biopsies– that were previously done under light microscopy. In consequence, the interest in building clinical decision support systems utilizing the data derived from whole-slide images arose. Within the article, descriptions of the first clinical-grade decision support system in pathology were described. Utilizing a vast dataset of nearly 50,000 images pulled from 15,000 cancer patients, researchers were able to create an effective machine learning algorithm which detected prostate, skin and breast cancer with near-perfect accuracy. Published in Nature Medicine, the research highlighted within the article helped underscore the potential such computational algorithms have in serving as clinical decision support systems for pathologists with challenging classifications or needs for quantifying subjective staining. More importantly however, the article highlights the critical needs in generating large datasets of trainable images derived from several hospitals and countries to improve generalization of systems.  In turn, moving forward, it will be critical that pathologists and informaticians collaborate in order create effective decision support systems that can be implemented at any site.

Considering the viewpoint of a Chief Information Officer or Digital Pathologist major promises in the application of such technologies and considerations of the design of such support systems can be made. With the integration of molecular testing and next-generation sequencing in accurately classifying cancers, challenges arose for poorer nations and rural hospitals that were ill-equipped to make the necessary technology purchases to accommodate such changes. In consequence, such hospitals were required to outsource cancer cases to larger hospitals– effecting treatment speed and convenience for patients. Computational clinical decision support systems in pathology however may alleviate these challenges by providing support in pathological diagnosis based upon histological images that can be preformed at any hospital. However, with such systems requires considerations like those highlighted by the article. While a single site may see hundreds of cancer patients and obtain thousands of images, biases in the data may occur from the populations seen at a hospital, differences in the processing of histological stains and scanning and entry of clinical data. Thus, training of computational systems across several sites is critical. However, the standardization of data entry and pre-planning of creating data-use agreements will be equally important. While computational models can be robustly trained to learn the differences that exist between sites, lack of standards can introduce wide data variability that weakens model training. Moreover, data exchange often can be time consuming and requires advanced pre-planning to ensure development moves forward.

Public Health Impact

With the rise of digital pathology and decision support systems that aid diagnosis, major value to the public health system arises in aiding undeserved regions and during periods of crisis. As a public health official, a major application of these tools can be considered during a period of crisis like Covid-19. With a pathogen with high infection risks, a public health official is poised to make recommendations to physicians that best ensures clinical care is maintained but does not put care providers at risk or potentially cause further disease spread. With digital pathology, remote diagnosis can be made at home. Moreover, with clinical decision support the time to diagnosis is improved. During such a scenario, a public health official can more effectively mobilize a sector of physicians to remain remote to ensure safety, but also utilize digital pathology to maintain care for patients that need it. Moreover, in smaller rural communities where disease is highly prevalent or hospitals cannot keep up with intake, digital pathology platforms create a mode that allows distant collaboration with physicians at alternate sites. While the news piece highlights digital pathology in the context of cancer, pathologists play roles in several organ systems including lung health. Thus a public health official challenged with mobilizing workforce to areas of Covid spread where rapid care in lung diagnostics is needed, such digital tools can allow a pathologist in Ohio to provide expertise to a rural site in Kansas– relaxing aid distribution.

Timeline: Clinical Decision and Computation– A History of the Integration of Machine to Clinical Decision Support

Link to timeline