OPINION
The future of digital pathology: Why mimicking radiology will hinder progress
The Advanced Research Projects Agency for Health’s (ARPA-H) ImagiNg Data EXchange (INDEX) programme and the Digital Pathology Association (DPA) have established a thorough framework for digital development in pathology, while at the same time acknowledging that infrastructure challenges are hindering pathology AI advancements. However, do they bear some responsibility for this? The common criticism of pathology’s
slow progress with digital technology (often due to misguided comparisons with other diagnostic disciplines) is incorrect. Attempting to align pathology with radiology and cardiology’s trajectory will not only impede the success of its digital roadmap but also compromise efforts to address its significant infrastructure challenges.
A singular origin Pathology’s unique infrastructure necessitates a tailored approach. 1. Different mediums. Unlike cardiology and radiology, pathology’s centuries- long reliance on glass slides means it faces a significantly longer and more challenging path to digitalisation.
2. Scale. Pathology images are substantially larger than those in other disciplines. While exact sizes vary, a high-resolution pathology image can be 80 times the size of a radiology image, clearly explaining the immense gap between current IT infrastructure and what’s actually needed. Furthermore, to extract adequate diagnostic insights, whole slide images (WSIs) must also be stained. This translates to considerably larger datasets per patient, with WSIs with 10 stains requiring over 100Gb per patient. This compares to only 60Gb per patient for a whole-body CT scan, and around 250Mb per patient for an echocardiographic study
3. Uncommon operations. Pathology’s shift is also uniquely complex, largely due to its integrated Laboratory Information Systems (LIS), which manage everything from the specimen lifecycle to result verification. This contrasts with radiology’s more segregated approach, where Radiology Information Systems (RIS) handle workflows and reports, and Picture Archiving and Communication Systems (PACS) manage the images themselves.
Infrastructure hurdles While ARPA-H INDEX and the DPA recognise that fragmented and underdeveloped digital pathology infrastructure significantly hinders its rollout and clinical application, it provides little actionable guidance on how to resolve this bottleneck.
Pathology is the biggest consumer of digital storage by a significant margin. Industry analysis indicates that a large hospital will likely require at least a petabyte (Pb) – equivalent to 1,000 terrabytes (Tb) - of storage annually to meet its digital pathology needs. This creates an enormous demand for storage solutions that are compliant, scalable, and high performing.
Because data is fragmented and file
formats are inconsistent, data exchange is hindered, workflows are slowed, and communication between departments becomes more complex.
The sheer size of WSI files necessitates considerable network bandwidth and robust, high-speed connectivity. However, many are ill-prepared for such vast data volumes, leading to frustratingly slow transfer speeds and frequent service interruptions.
Paving the way Despite the allure of digital pathology, we must realistically assess if our underlying infrastructure is ready to support it at scale. If healthcare network infrastructure continues to hinder digital pathology, following in the path of radiology and cardiology is futile. Our focus should shift to re-evaluating our direction and investing in resolving these infrastructure challenges, rather than joining an unwinnable race. n Interdepartmental collaboration. Advancements hinge on interdepartmental collaboration to share infrastructure and budgets to ease the storage burden. Integrating with existing DICOM systems is also crucial for standardised image formats and interoperability, future-proofing pathology data within a unified system for universal benefit.
n Modified operations. Adopting digital workflows means altering ingrained laboratory practices, making their buy-in crucial. Key obstacles such as a lack of real-time magnification control also need to be overcome. This requires pathology diagnostic
WWW.PATHOLOGYINPRACTICE.COM AUGUST 2025 About Liam Canavan
Liam Canavan is Healthcare Lead at Loadbalancer. org. He partners closely with medical imaging providers and hospitals across three continents. He champions strategic alliances and cutting- edge technology to tackle complex data storage challenges and ultimately enhance patient care worldwide. Loadbalancer’s application delivery
solutions are vital for over 1,000 healthcare networks globally, supporting 200 million- plus studies annually. The company’s extensive reach is demonstrated by partnerships with a quarter of UK NHS Trusts, the NIH and VISNS Network in the USA, and Saudi Arabia’s MOH, alongside most Fortune 100 Enterprise Imaging companies. Its solutions are also validated and integrated with major vendors like Philips, Fujifilm, and GE HealthCare.
processes to be completely redesigned to ensure accuracy and mitigate these disadvantages.
n Data compression. Digital pathology, heavily reliant on WSI, demands a robust network; high-throughput scanners need 1 Gb/second to database servers and 10 Gb/second to storage. To handle this, data compression is crucial. While lossy compression offers smaller files, the potential absence of diagnostic information makes lossless compression the preferred option, though even this isn’t a complete solution.
A singular path
The significant digital infrastructure demanded by comprehensive WSI implementation presents a unique hurdle for pathology. At present, the intense focus on AI is, in fact, impeding progress. While a slow, deliberate infrastructure build-out might lack the glamour of AI breakthroughs, empowering pathology with the space and resources to develop its digital imaging infrastructure from the ground up will ultimately create a more stable and beneficial ecosystem for everyone, including future AI advancements.
Liam Canavan
Loadbalancer.org
5
Page 1 |
Page 2 |
Page 3 |
Page 4 |
Page 5 |
Page 6 |
Page 7 |
Page 8 |
Page 9 |
Page 10 |
Page 11 |
Page 12 |
Page 13 |
Page 14 |
Page 15 |
Page 16 |
Page 17 |
Page 18 |
Page 19 |
Page 20 |
Page 21 |
Page 22 |
Page 23 |
Page 24 |
Page 25 |
Page 26 |
Page 27 |
Page 28 |
Page 29 |
Page 30 |
Page 31 |
Page 32 |
Page 33 |
Page 34 |
Page 35 |
Page 36 |
Page 37 |
Page 38 |
Page 39 |
Page 40 |
Page 41 |
Page 42 |
Page 43 |
Page 44 |
Page 45 |
Page 46 |
Page 47 |
Page 48 |
Page 49 |
Page 50 |
Page 51 |
Page 52 |
Page 53 |
Page 54 |
Page 55 |
Page 56 |
Page 57 |
Page 58 |
Page 59 |
Page 60 |
Page 61 |
Page 62 |
Page 63 |
Page 64 |
Page 65 |
Page 66 |
Page 67 |
Page 68 |
Page 69 |
Page 70 |
Page 71 |
Page 72
