Scan Times and Hidden Costs
The challenges of using digital pathology in a lab environment
Prasanth Perugupalli | | 2 min read | Opinion
Scanning is a vital part of the digital pathology pipeline. But what are the hidden pitfalls that stymie its success? And what are the cost implications? Find out in part two of our six-part “Barriers to Adopting Digital Pathology” series.
As pathology labs evaluate their options for adopting digital workflows, the most critical decision comes down to the total cost of this new adventure.
The workflow efficiencies that digital pathology can bring to planning, scheduling, and physician engagement are indisputably attractive. There is increasing chatter about AI-driven digital labs of the future, and labs want to deploy digital pathology – albeit at a small scale – so that they can “talk the talk” and plan their strategies for the future.
The science of making images is fundamental to digital transformation. Glass slides are converted into digital whole slide images, which can be stored, uploaded, and transported over the internet. A tremendous focus falls on the time and effort to make this digital data – and rightly so. One metric that is most misrepresented in the industry is scanning time. Vendor specs may not include the pre-analytics and prep time for the slides before they are loaded into the scanners, including checking for a misplaced coverslip or a label – or the time it takes to load the slide baskets – laborious tasks technicians silently dread.
Indeed, significant resources can go into making decisions on how to scan a particular specimen for first-pass success. Manual quality checks add at least one to two minutes per slide and will result in a call for a rescan if errors are found. Each of these steps add costs, mostly in the form of trained manpower. At what point does the total cost of digital adoption become prohibitive? Put another way, a great digitization solution is one that mitigates the need for pre- and post-scan intervention.
Scanning issues are exacerbated when images are fed into an AI pipeline. Labs may need to invest in additional QA tools as a precursor to the AI because algorithms have yet to achieve the low-level general intelligence to overlook certain bad patches on images.
Lab managers should model total cost of operation, hiring needs, material flow, and overall ability to go digital in their daily practice. Several early adopters have deployed excess capacity through the addition of more scanners, which does not seem sustainable.
A “Digital Pathology as a Service” business model is an option that puts the onus on the solution providers to help the lab managers achieve the optimal deployment plan. An analogy exists in the mobile phone industry, which realized at the onset of 3G services two decades ago that data-driven paradigms are complex. It was only when capacity planning and operation were left to the system providers, that maximum efficiencies in operation costs were achieved.