Editorial Computational Pathology An Emerging Definition David N. Louis, MD; Georg K. Gerber, MD, PhD; Jason M. Baron, MD; Lyn Bry, MD, PhD; Anand S. Dighe, MD, PhD; Gad Getz, PhD; John M. Higgins, MD; Frank C. Kuo, MD, PhD; William J. Lane, MD, PhD; James S. Michaelson, PhD; Long P. Le, MD, PhD; Craig H. Mermel, MD, PhD; John R. Gilbertson, MD; Jeffrey A. Golden, MD A dvances in high-throughput laboratory and health information technologies are revolutionizing the dis- ciplines of pathology and laboratory medicine. The ability to extract clinically actionable knowledge using computational methods from complex, high-dimensional laboratory and clinical (digital) data, thereby yielding more precise diag- noses, disease stratification, and selection of patient-specific treatments, will clearly be a significant and important realization in the delivery of health care. Pathologists, who are at the nexus of diagnostic data, models of disease pathogenesis, and clinical correlation, are ideally positioned to provide leadership in the emerging ‘‘big data’’ era of medical care. We thus propose a vision for a new discipline of computational pathology. We define computational pathology as an approach to diagnosis that incorporates multiple sources of raw data (eg, clinical electronic medical records; laboratory data, including ‘‘-omics’’; and imaging); extracts biologically and clinically relevant information from those data; uses mathematical models at the levels of molecules, individuals, and populations to generate diagnostic inferences and predic- tions; and presents that clinically actionable knowledge to customers through dynamic and integrated reports and interfaces, enabling physicians, patients, laboratory person- nel, and other health care system stakeholders to make the best possible medical decisions (Figure). This vision goes beyond an information technology or informatics-centric view and leverages the core competency of pathology—the understanding of disease processes at the molecular, individual, and population levels, and the ability to effectively integrate and communicate clinically actionable knowledge. Realization of this vision will require changes in the practice of pathology, which is currently an order-driven, observational discipline. Our clinical laboratories generate more observations than any individual can reasonably interpret for clinical care. Although the data we provide to clinicians are largely reported as unique and independent results, the number of potentially meaningful relationships among combinations of observations is astronomic, and the workup (the process and order of the observations) has become increasingly complex and expensive. The under- standing and high-level interpretations of those relation- ships remain in the mind of the pathologist and associated clinician, with the knowledge gap constantly increasing, which is not a sustainable situation. Four driving factors have now created opportunities to move the field forward: (1) pathology data are increasingly digital and remain the most detailed and structured sets of information data in patient health records; (2) laboratory information systems are increasingly powerful, flexible, and integrated, allowing more complex analyses and interfaces with other diagnostic systems; (3) pathologists have access to the entire (digital) clinical record of patients, which gives them the ability to correlate laboratory data with clinical status and endpoints to develop foundational medical and biological knowledge; and (4) large-scale, clinical, pheno- typing data are increasingly being collected and stored in structured databases, which enhance the capacity of pathologists to query and integrate information across many subjects to drive population-based analyses. By developing the tools to harness those drivers, pathology can improve delivery of medically actionable knowledge over time and across populations and increase the efficiency of health care delivery. In this manner, the discipline of pathology can move from observation alone to a combination of observation, truly integrative interpreta- tion, and longitudinal workup of patients. Although the parts of computational pathology exist, the whole does not. Thus, below, we describe those components and the people and infrastructure within the specialty that are needed to bring the parts together into a coherent whole to realize the vision of computational pathology. Accepted for publication February 3, 2014. From the Partners Program in Computational Pathology and the Department of Pathology, Massachusetts General Hospital, Boston (Drs Louis, Baron, Dighe, Getz, Higgins, Michaelson, Le, Mermel, and Gilbertson); the Department of Pathology, Brigham and Women’s Hospital, Boston (Drs Gerber, Bry, Kuo, Lane, and Golden); the Departments of Pathology (Drs Louis, Gerber, Baron, Bry, Dighe, Getz, Kuo, Lane, Michaelson, Le, Mermel, Gilbertson, and Golden) and Systems Biology (Dr Higgins), Harvard Medical School, Boston; and the Broad Institute, Cambridge, Massachusetts (Drs Getz, Le, and Mermel). Drs Louis, Gilbertson, and Golden are co-senior authors. The authors have no relevant financial interest in the products or companies described in this article. doi: 10.5858/arpa.2014-0034-ED Reprints: Jeffrey A. Golden, MD, Department of Pathology, Brigham & Women’s Hospital, 75 Francis St, Boston MA 02115 (e- mail: jagolden@partners.org). Arch Pathol Lab Med—Vol 138, September 2014 Editorial—Louis et al 1133