Digital Pathology is a new data source to transform medicineMedscape – Oct 15, 2014.
This was the hook line of a 2014 article in Medscape we stumbled across the other day. Ten years later, it’s time to take stock and to check if the impact of digital pathology was indeed transformational. (Spoiler alert: They were right!)
Here’s our ultimate guide:
10 Things You Should Know about Digital Pathology –
and things to consider for the bright digital future ahead
|The Covid pandemic helped: The possibility to work from home has become one of
the key drivers for digital pathology (and telepathology).
|And so has the lack of new pathologists (“workforce crisis”). A new hire might prefer
working in a (digital) practice that supports occasionally working from home over a
traditional (analog) practice.
|Finally, in 2022 and 2023 a multitude of AIs received CE-IVDR clearance and can be
used in the routine.
|Digital pathology is often referred to as the “3rd revolution” in pathology, following
the introduction of immuno-histochemistry (IHC) as the 1st revolution and molecular
pathology as the 2nd revolution.
|Digital pathology produces gigabytes of data – each day.
A single scan takes between 1-3 GB, sometimes even more.
Most labs that decide to go fully digital and scan each
and every glass slide use a “rolling archive” that starts overwriting scans after 1-3
months, ensuring the archive does not grow endlessly. Slides can be tagged to keep
them forever, e.g., when they are medically interesting or may become relevant
from a legal perspective. Some labs store slides “on-prem” in a local storage
(higher up- front costs and hardware has to be replaced during its lifetime),
others in the cloud (continuous metered fees per terabyte).
|Main reasons for NOT going digital are:
1. The workflow is changed entirely: Technical assistants constantly feed slides into
the scanner. Scanners can work over night. Pathologists open and review case on
their computer. Scanners represent an additional processing step and with it an
additional point of failure. Scans can sometimes be out-of-focus, and slides may
break, which can even jam the scanner.
2. The costs: Approximately one scanner per pathologist is required. Additionally, an
Image Management System (IMS, aka Histo-PACS) has to be introduced and it
needs to be compatible with the LIS to ensure the same case is opened in
both systems. LIS and IMS are typically open side-by-side on two monitors.
Some labs use this occasion to switch their LIS as well. Aside from the up-front
costs, the archive for storing the scans produces high costs over time.
The business case is not clear.
The image quality: Some pathologists prefer the quality of their analog microscope
over that of a scanner.
3. The image quality: Some pathologists prefer the quality of their analog
microscope over that of a scanner.
|There is still no standard imaging-format. While in radiology, it is crystal clear that
images are stored in the DICOM format, in pathology most vendors use their own
proprietary formats. Sure, a DICOM-WSI format also exists and is gaining traction.
Maybe one day it will replace the proprietary formats.
Let’s check back in another 10 years.
|The little brother of digital pathology is to introduce computer-readable slide labels
that contain barcodes or QR-codes, coding the case ID and type of stain.
|Digital pathology is a pre-requirement for Computational Pathology,
i.e., going digital unlocks the door into the world of AI.
AIs can be integrated into the IMS to automatically conduct analyses
after scanning – before a pathologist even opens the case.
For instance, introducing an AI that detects metastases in lymph node sections
allows sorting the slides of a case by decreasing tumor presence probabilities;
heatmaps can guide the pathologist’s gaze to a particular area in a slide. This can save
time, or it can be regarded as a second look and decrease the risk that a (micro)
metastasis is missed. Other types of AIs include Gleason Grading for prostate cancer,
Ki67 or PD-L1 scoring for breast or NSCLC. Many more AIs are available and new
ones are becoming available at a rapidly growing pace.
|Currently, in many countries, labs do not get reimbursed for the additional cost for
scanning or AI analysis. An exception is the US, where scanning is reimbursed.