Pathology RE-Imagined: The History of Digital Radiology and the Future of Anatomic Pathology

The clinical practice of anatomic pathology and radiology are similar at several levels. Both disciplines require an imaging modality to collect primary data. For radiology, the modality can be varied but primarily detects photons, magnetic field perturbations, or ultrasonic waves, which are reconstructed to represent the original anatomy. For pathology, photons generated from a microscope are reconstructed in the brain to embody the original morphology. In both disciplines, the overwhelming majority of diagnoses are made by assessment of structural features. In pathology, this is called morphology, but the principle is identical to radiology that measures anatomy. In radiology, additional information is analyzed for a minority of cases using “molecular techniques” such as fluorodeoxyglucose positron emission tomography or other molecular imaging techniques. Likewise, in pathology, immunostains, nucleic acid hybridization, or various enzymatic special stains are used to examine genes, proteins, or enzymes. However, this is also in a minority of cases. In both disciplines, molecular interrogation is not fully quantitative, and physicians generally struggle with a need to improve quantitation.A less obvious, but significant, similarity between radiology and pathology is the size of digital images. It is well known that digital radiology carries the burden of large image sizes, which has led to the advent of sophisticated picture archiving systems and associated standards that are pervasive within the modern hospital system. Interestingly, the average digital pathology sample (hematoxylin-eosin digitized at 0.5 μm per pixel resolution) is between 2 and 10 gigabytes. This is an order of magnitude larger than an average radiology examination. Given the volume of slides in a standard pathology laboratory, infrastructure to manage large images will need to exist in the clinical laboratory if digital pathology is to become a reality.An important similarity also exists when evaluating the value of digital solutions. It has been established in the radiology department that an “all digital environment” can significantly improve workflow, which translates into higher productivity and less cost. In pathology, the same opportunity exists. There are great workflow inefficiencies in the pathology department that result from the necessary physical connection between glass slides and the highly paid labor required to read them. The overall process to achieve this connection requires labor to collate glass slides with the proper requisitions and prepare them for physical sign-out as well as recollate them after reading and preparing for storage. In cases in which consults are needed, glass slides need to be recollated, packaged, shipped, unpackaged, collated, signed out, read, reported into a separate system, repackaged, and shipped back to the originating institution. In cases in which secondary reads are needed, the cases are requested and time and labor are spent retrieving them. Gathering cases and data for tumor boards and/or educational presentations is also a huge labor burden, during which time standard primary cases are not being read resulting in significant opportunity costs. Lastly, in multi-institutional health care systems, balancing the load between the primary laboratory and satellite sites is inefficient resulting in a buildup of slides at one site and waiting for glass at another site. The entirety of this inefficiency can be eliminated if the primary data (ie, glass slides and the microscope needed to read them) were disconnected from the person reading the data. History has shown this to be true in radiology.A digital environment is not just about the process of diagnosis but the quality of diagnosis as well. This is perhaps the most significant advantage of digital data. You cannot hang a multislice computed tomography examination on a light box. Similarly, you cannot perform three-dimensional imaging of tissue or multiprotein colocalizations and full dynamic quantitation by looking through a microscope at a single static image. There is no argument that digital technology in radiology has resulted in earlier and enhanced diagnostic accuracy. The poster child for this is volumetric computed tomography and multimodality fusion, which has made great strides in detecting disease earlier and pinpointing its origin to allow for earlier and more targeted therapeutic interventions.1 Is there any reason that this would not occur in pathology as well? In the clinical research literature, we are already seeing examples of automated subcellular compartmentalization of proteins in tissue sections, which have significant power to stratify patients beyond traditional microscopic review.23 Even in the absence of such advances, the simple ease of gathering a second opinion or informally gathering a consensus on a difficult hematoxylin-eosin would undoubtedly enhance diagnostic accuracy and ultimately positively affect patient care. Thus, if radiology is used as a guidepost, the future of anatomic pathology will be the movement from glass to digital, from semiquantitative to full quantitation, and from a single biomarker to multidimensional data.Of course, there are reasons why the adoption of digital pathology has not been taken too seriously and these reasons can generally be categorized into 3 general areas: (1) the technology is too slow to scan the total volume of slides; (2) the cost is too expensive, particularly storage; and (3) the value is not there to justify the expense. Although all of these statements were largely true as recently as 5 years ago, they are less true today. First, many vendors are working toward a high throughput scanner, which is capable of scanning whole slides in 30 seconds or less at high quality with autoloaders of more than 200 slides. Scanning slides quickly is not against the laws of physics. Second, the cost of storage continues to plummet and the speed of processors continues to grow. This is perhaps the single greatest change that will enable digital adoption in pathology. Lastly, value can be easily extracted from workflow/efficiency improvements and an eventual increase in diagnostic accuracy, as has occurred in radiology.There are still major obstacles that will keep adoption from becoming the norm. Although great progress has been made on scanners, storage, and browsers, there are critical missing pieces that are needed to capture true value; namely, an infrastructure whose mission is to capture efficiency in the overall pathology workflow. A foundation that includes accession, scheduling, load balancing centralized image retrieval, seamless access to a laboratory information management system, reporting, secondary requests, consult requests, billing, and links to the hospital information systems are all needed for the value to be realized (Figure). Without this, it is very unlikely the cost-benefit equation will tip in favor of an all-digital workflow in anatomic pathology. This is even more critical when we realize that the “filmless” benefit that seeded early adoption in radiology does not exist in pathology. Primary data, at least for the foreseeable future, will be on glass. Thus, the initial value tied into workflow becomes even more critical for adoption. Another barrier is hospital economics that are not accustomed to large capital investments in anatomic pathology in the absence of a proven cost-benefit equation.In summary, there are great similarities between radiology and pathology, and an all-digital solution can improve workflow and diagnosis in the anatomic pathology market as has been demonstrated in radiology. However, adoption will be slow and dictated by the benefits of an end-to-end solution that includes clinical workflow and connection to the hospital information systems infrastructure.