TalkingPoint automation supports digital pathology AI triage
The introduction of digital pathology and other new technologies in recent years means that it is now possible for pathologists to use digital images and artificial intelligence (AI) systems to triage cases.
Managing a caseload of approximately 35,000 histopathology cases every year, Southwest Pathology Service (SPS) is an integrated pathology service covering Musgrove Park Hospital, Taunton (MPH) and Yeovil District Hospital, Yeovil (YDH).
In Taunton, the pathology service has been an early adopter of digital pathology, commissioning digital slide scanners more than a year ago.
With the adoption of any new digital healthcare technology, there are often issues around interfacing with existing systems. In hospital pathology laboratories, it is also not unusual for there to be several platforms in use with little interoperability between them.
Dr Fred Mayall, Consultant Cellular Pathologist, Somerset NHS Foundation Trust, explains: “The problem we had was that our digital microscope slide scanning platform would not easily integrate with our other laboratory systems. Our research group had developed cutting-edge AI analysis software but we still had to find a way to export digital images to allow them to be processed by the AI. The solution offered by the vendor of the scanner was a manual method that was slow, and no suitable for processing large volumes of slides.”
The Pathology service at Taunton has been using TalkingPoint’s workflow automation software for several years to provide other process automation in the department, so it was a natural step to consider using TalkingPoint to provide a solution.
Dr Mayall explains: “The process that TalkingPoint now drives takes images from the scanner and passes them to the artificial intelligence engine. The AI can sort biopsies of the large intestine so that abnormal cases can be identified and prioritised for reporting.”
Downloading the images used to be time consuming as they had to be done one at a time. Users had to load the case, identify the specific slide required, select to export the file, wait for it to export and then save it – with each slide taking several minutes. In context, a typical gastrointestinal biopsy case may generate 10 or more slides that would take at least 20 minutes to download and save.
TalkingPoint’s automation (the interface for which is shown below) emulates the activity of an expert user of the platform to do repetitive tasks at high speed.
Dr Mayall continues: “It would take approximately 90 minutes a day for a user to import the slides into the AI engine from the digital scanner; a process requiring repetitive case number entry, mouse clicks and long pauses, with constant vigilance. With TalkingPoint, the process can be left unattended – for example, we could leave it to complete overnight. The automation is robust. Importing the slides had become an onerous task whereas this task is now managed entirely by the TalkingPoint automation!”
Malcolm Grant, Managing Director, TalkingPoint added: “This was a tricky automation to build as the automation would be driving browser-based software. After a successful initial test to check that TalkingPoint could control the software, an extremely robust workflow was built so that it would run unsupervised. For example, if something in the process didn’t work, the workflow would find out why and then try again automatically.
The resultant TalkingPoint automation now reads a file with all the required cases and drives the export process automatically, placing the files into an image folder.
This is a good example of how process automation can deliver efficiency benefits and aid the adoption and integration of new and developing technologies within Pathology and other healthcare specialties.”