Approximately 1.4 million women per year depend on pathologists to accurately interpret breast biopsies for a diagnosis of benign disease or cancer. Diagnostic errors are alarmingly frequent and likely lead to altered patient treatment, especially at the thresholds of atypical hyperplasia and ductal carcinoma in situ, where up to 50% of cases are misclassified. The causes underlying these errors remain largely unknown. Technology similar to Google Maps now allows pan and zoom manipulation of high-resolution digital images of glass microscope slides. This technology has virtually replaced the microscope in medical schools and is rapidly diffusing into U.S. pathology practices. No research has evaluated the accuracy and efficiency of pathologists' interpretion of digital images vs. glass slides. However, these digital slides offer a novel opportunity to study the accuracy, efficiency, and viewing behavior of a large number of pathologists as they manipulate and interpret images. The innovative analytic techniques proposed in this application are similar to those used to improve the performance of pilots and air traffic controllers.
Our specific aims are:
Aim 1. Digital Image vs. Glass Slides: To compare the interpretive accuracy of pathologists viewing digitized slide images over the Internet to their performance viewing original glass slides under a microscope. A randomized national sample of pathologists (N=200) will interpret 240 test cases in one or both formats in two phases. Measures will include a diagnostic assessment for each test case and for digital slides, cursor- (i.e., mouse) tracking data and region of interest (ROI) markings. Completion of this aim will establish benchmarks for the comparative diagnostic accuracy of whole-slide digital images.
Aim 2. Interpretive Screening Behavior: To identify visual scanning patterns associated with diagnostic accuracy and efficiency. Detailed simultaneous eye-tracking and cursor-tracking data will be collected on 60 additional pathologists while they interpret digital slides to complement data from Aim 1. Viewing patterns will be analyzed from computer representations of raw movement data. Videos depicting accurate, efficient visual scanning and cursor movement will be valuable tools in educating the next generation of digital pathologists.
Aim 3. Image Analyses: To examine and classify the image characteristics (including color, texture, and structure) of ROIs captured in Aims 1 and 2. Computer-based statistical learning techniques will be used to identify image characteristics that lead to correct and incorrect diagnoses. Characteristics of both diagnostic and distracting ROIs will be identified, linking all three aims. In summary, we will determine whether digitized whole-slide images are diagnostically equivalent to original glass slides. Our in-depth scientific evaluation of viewing patterns and characteristics of ROIs identified by pathologists will be critical to understanding diagnostic errors and sources of distraction. Optimization of viewing techniques will improve diagnostic performance and thus, the quality of clinical care.

Public Health Relevance

Pathologist errors in the interpretation of biopsy specimens can result in significant harm to patients. Digital imaging technology is rapidly diffusing into medical settings, providing a unique opportunity to obtain data on the process used by pathologists when interpreting slides. The results will provide the foundation needed to improve interpretive accuracy and efficiency, support quality improvement efforts, and ultimately reduce the burden of misdiagnosis on patients and the health care delivery system.

National Institute of Health (NIH)
National Cancer Institute (NCI)
Research Project (R01)
Project #
Application #
Study Section
Health Services Organization and Delivery Study Section (HSOD)
Program Officer
Demoor, Janet S
Project Start
Project End
Budget Start
Budget End
Support Year
Fiscal Year
Total Cost
Indirect Cost
University of Washington
Internal Medicine/Medicine
Schools of Medicine
United States
Zip Code
Mercan, Ezgi; Shapiro, Linda G; BrunyƩ, Tad T et al. (2018) Characterizing Diagnostic Search Patterns in Digital Breast Pathology: Scanners and Drillers. J Digit Imaging 31:32-41
Mercan, Caner; Aksoy, Selim; Mercan, Ezgi et al. (2018) Multi-Instance Multi-Label Learning for Multi-Class Classification of Whole Slide Breast Histopathology Images. IEEE Trans Med Imaging 37:316-325
Carney, Patricia A; Frederick, Paul D; Reisch, Lisa M et al. (2018) Complexities of perceived and actual performance in pathology interpretation: A comparison of cutaneous melanocytic skin and breast interpretations. J Cutan Pathol 45:478-490
BrunyƩ, Tad T; Mercan, Ezgi; Weaver, Donald L et al. (2017) Accuracy is in the eyes of the pathologist: The visual interpretive process and diagnostic accuracy with digital whole slide images. J Biomed Inform 66:171-179
Geller, Berta M; Nelson, Heidi D; Weaver, Donald L et al. (2017) Characteristics associated with requests by pathologists for second opinions on breast biopsies. J Clin Pathol 70:947-953
Onega, Tracy; Weaver, Donald L; Frederick, Paul D et al. (2017) The diagnostic challenge of low-grade ductal carcinoma in situ. Eur J Cancer 80:39-47
Samples, Laura S; Rendi, Mara H; Frederick, Paul D et al. (2017) Surgical implications and variability in the use of the flat epithelial atypia diagnosis on breast biopsy specimens. Breast 34:34-43
Jackson, Sara L; Frederick, Paul D; Pepe, Margaret S et al. (2017) Diagnostic Reproducibility: What Happens When the Same Pathologist Interprets the Same Breast Biopsy Specimen at Two Points in Time? Ann Surg Oncol 24:1234-1241
Lee, Christoph I; Chen, Linda E; Elmore, Joann G (2017) Risk-based Breast Cancer Screening: Implications of Breast Density. Med Clin North Am 101:725-741
Frederick, Paul D; Nelson, Heidi D; Carney, Patricia A et al. (2017) The Influence of Disease Severity of Preceding Clinical Cases on Pathologists' Medical Decision Making. Med Decis Making 37:91-100

Showing the most recent 10 out of 23 publications