In this project, Saber Informatics was asked to implement a workflow management system for high-volume streams of pathology images in clinical research. Our client sends human tissue samples for analysis to partners who send back high-resolution pathology images of tissues at the cellular level. Software-aided algorithmic analysis of the images is performed in-house by a cross functional team of scientists. Reviewers perform their respective tasks in a sequence that may vary depending on context: varying degrees of complexity in tissue images will require tweaking analysis parameters and re-running image recognition software before delivering a final result.
In a typical study, up to a hundred pathology images would be received for analysis but not all at once, depending on how partners deliver them. When finally ready, images would undergo automated fusion ("stitching") into complete, much larger images, followed by statistical analysis on a compute cluster. After a review by clinical pathologists, some images would need to be manually re-analyzed with updated statistical algorithms due to failures or inapplicability of the original statistical analyses. With multiple studies current at any given time, the review team can easily lose track of where each study is in the workflow and most importantly, what caused the delay. Statistical image analysis takes minutes to hours and if not automated, will consume significant human operator time and resources.
A workflow management solution for tissue image analysis helps cross-functional teams get their work done much faster.
Working collaboratively with Client scientists, we designed and implemented an algorithm-driven workflow management solution that includes a calculation engine, a visual dashboard, and a background alerting engine. The calculation engine automates statistical analyses and logs the results. The dashboard lists all current and completed studies based on the information from the Client's image management database. In each study a high-level view of progression is shown so that the review team is always informed of where each study is in the process. Whenever a team member (e.g. statistical analyst) completes their review work, the system sends an alert to other team members (e.g. pathologists) who can immediately begin to perform their analysis tasks. Overall progress is always visible in the dashboard. Detailed stepwise logs with actions and staff IDs can also be accessed from the main dashboard page if needed.
Saber Informatics helped our Client improve team productivity by automating workflow orchestration for pathology image analysis. Any delays in image acquisition or processing are now clearly visible in the shared dashboard and can be addressed immediately. The review team can now complete their work much faster than before.
