For an efficient AI-powered diagnostic workflow, it is crucial to ensure that pathology images and associated metadata are connected at the source
David Dimond | | Longer Read
If we are to unlock the true potential of the digital pathology workflow, then the creation and interoperability of metadata is essential. Attaching relevant data to medical images allows consultants to collaborate on diagnoses across time and from multiple locations – and will likely open the doors to future developments in AI-driven insights.
After attending a recent digital pathology event, I became excited about the interoperability commitments industry vendors are making. Breakthroughs from standards-based working groups, such as Digital Imaging and Communications in Medicine (DICOM) and the Integrated Healthcare Enterprise (IHE), along with updated guidance from the US Food and Drug Administration, have set the stage for a more sustainable and innovative approach to digital pathology.
The recent increase of activity in the DICOM standards community, for example, is a clear sign of progress. The first meeting of DICOM Working Group 26 (DICOM WG-26) took place in 2005, but the scope of their Connectathon interoperability demonstrations at this year’s Pathology Visions conference reached new levels.
For the uninitiated, both the DICOM organization and IHE are standards development organizations, with the goal of improving the interoperability of health information technology (HIT) systems – addressing specific clinical needs in support of optimal patient care. The DICOM organization was established in 1983 to develop a specific standard for the communication of medical imaging information. DICOM WG-26 is a working group within the DICOM organization that focuses on the development of a DICOM imaging standard for whole-slide imaging (WSI) in digital pathology. IHE promotes the coordinated use of established standards to achieve interoperability between HIT systems and effective use of electronic health records (EHRs) – and has a strong relationship with DICOM. Pathology and laboratory medicine (PaLM) is a domain of the IHE that addresses information sharing and workflow related to in vitro diagnostic testing in anatomic pathology, clinical laboratories, and at the point of care. The standards and use cases have been defined and vendors are starting to adopt them – first in academic research organizations and then within traditional healthcare providers.
I’ve evangelized DICOM and IHE standards for Radiology and Cardiology PACS for years. I’ve built teams who used the standards in both advisory and consulting work to help select, design, and implement some of the largest PACS systems in the US. These experiences taught me to appreciate the nuances of technology integration in healthcare.
Although the increase in interest is exciting, there’s a key feature at the center of these developments that requires careful consideration. From my observations, what we need going forward are “metadata-enhanced workflows.”
Digital pathology has the potential to become the new standard of care, but digitization is only one part of the equation. Digital pathology’s biggest strength comes from metadata – the information that can be attached to digital images to become part of an efficient medical workflow. This includes everything relevant to the images: annotations, medical notes, patient identification, clinical order data, modality protocols, and date and time stamps. It is the composite of all information that is associated with a medical image, whether collected at the time of imaging or added after the fact, to create a “packet” needed for workflow enablement, data integrity, provenance, and security.
Without metadata, an archive of WSI files is like storing files on a hard drive using just file and folder names – cumbersome to categorize, search, and use. Conversely, if the pathology images come with embedded data and (more importantly) DICOM-standardized tags, there is a greater opportunity to generate insight during the clinical diagnostic workflow and research phases. Subsequent patient care also becomes interconnected as the metadata builds, ensuring that each WSI file retains an audit trail and giving diagnosticians and clinicians access to a holistic, longitudinal patient view. Researchers also benefit because metadata enhances their ability to harvest high-quality data from biobanks and distributed research archives.
The ability to “line up” the WSI tags with diagnostic images, case histories, and electronic health record (EHR) data also allows pathologists to effectively collaborate with virtual teams, such as tumor boards, for difficult-to-diagnose cases. This has the potential to significantly impact patient survival rates because consultations happen rapidly and collectively. Furthermore, peer reviews and multidisciplinary meetings can become more seamless. Considering the global shortage of pathologists, streamlining their workflow will yield an enormous efficiency benefit, helping them use their valuable time more effectively.
Putting the pieces together
Combining data with images is a complex task that relies heavily on improved data interoperability. When radiologists made the switch to digital, for instance, it took most radiology departments up to 15 years to realize the benefits of enterprise Picture Archiving and Communication Systems (PACS). The prior lack of interoperability inhibited productivity, slowing down adoption rates. The good news is that, in comparison, progress in digital pathology technology is happening at light speed – largely thanks to the groundwork radiology has laid. For example, DICOM Working Group 26 is leveraging DICOM standards by combining WSI images and patient data into one format. It’s essential to have a standard like this so that equipment and software from different vendors can interoperate. DICOM Working Group 26 is also collaborating with teams from the IHE initiative in pathology and laboratory medicine to define how data relating to specimens, diagnostic observations, and documentation should be structured.
For pathologists to take full advantage of digital pathology, they must look at it from a workflow-first perspective and determine what can be improved when moving to a digital workflow. The ideal digital workflow includes access to all case-related whole-slide images via an image viewer integrated with the laboratory information system (LIS). Pathologists can then capture regions of interest – preferably using some automated capabilities such as barcodes and other forms of metadata – and seamlessly export these to the LIS report as needed. They should then be able to share this data with colleagues around the world for secondary consults, using vendor-agnostic communication protocols for improved collaboration. One of the key goals is to invest in an interoperable digital pathology solution, avoiding the pitfalls of potentially obsolete proprietary systems in the future – a problem currently seen with some of the post-PACS radiology Vendor Neutral Archive (VNA) implementations.
I currently work with medical institutions to develop and implement this type of workflow. By partnering with leading digital pathology scanner and PACS vendors, we gather bundled solutions to enable data portability and accessibility. For example, we provide the technology and infrastructure solutions to enable one vendor’s unique indexing technologies, which adds metadata in the form of hardware-generated barcodes at the point of scanning to link the relevant data to the patient EHR from the outset. We did this by developing specialized storage solutions to maintain the link between images and metadata to span multiple public and private clouds in various geographic locations. By implementing this approach, healthcare organizations not only gain the data mobility and accessibility needed, but also avoid the “data gravity” problem, in which the sheer size of data impedes its use outside its original repository.
As the quantity of digital pathology data grows, increasingly sophisticated analytical platforms are needed to glean new insights, opening doors to machine learning and artificial intelligence (AI). There is growing awareness – particularly from IT customers – that there’s a wealth of future benefits in store for patient treatment when insights from digital pathology images and data can be combined with other disciplines like genomics and radiology.
By systematically tagging WSI files with metadata, organizations can categorize whole-slide images by patient name, diagnostic site, institution name, and more. They can integrate images with patient reports by cross-referencing demographic data, automatically control access to clinical data based on level of network access, label images for rules-based data retention purposes, and generate end-to-end custodial audit trails from the moment of making the scan. The resultant metadata-enhanced digital pathology files allow healthcare and life sciences organizations to enhance clinical collaboration, streamline reporting, strengthen patient data security, and simplify data management.
Within this highly specialized area of digital pathology, research organizations and industry must work together on solutions that incorporate the standards referenced in this article. By doing so, they will be able to share pathology data seamlessly with other departments, both internally and with other healthcare and life sciences organizations. Once the integration of disparate systems with well-formulated and standardized data tags has been established, digital pathology can take the next step toward an enhanced workflow across the continuum of care. From there, a whole new world of AI-driven, federated data analytics tools will present opportunities that we’ve only just begun to imagine.