We used a Dell Precision M4800 laptop and a SensoMotoric Instruments (Boston, MA) Remote Eye Tracking Device (RED; 250Hz, ≤ 0.5˚ gaze position accuracy) attached to a 22” color-calibrated Dell liquid crystal display (LCD) monitor (at 1920 × 1080 resolution). Images were displayed on a custom digital viewer software using the DeepZoom Silverlight application (Microsoft, Inc.) in the web browser. The viewer tool allowed participating pathologists to zoom (1× to 60×) and pan the digital whole slide images while maintaining high resolution, and make measurements and annotations. While a pathologist used the viewer tool, data were automatically logged (at approximately 10Hz) to a local SQL database, to include zoom, position, and annotation data.
To collect diagnostic information from participants after reviewing each image, we developed a histology form on the Qualtrics web-based platform; histology forms are commonly used by pathologists to record their observations and interpretations of images [31 (link), 41 (link)]. The histology form asked participating pathologists to provide the single most advanced diagnosis by selecting among the five diagnostic categories detailed above (ranging from benign to invasive carcinoma). They also rated the difficulty of interpreting the image and their confidence with their determination (both on scales from 1–7). Of less relevance to this paper, information was also gathered about histopathological features (e.g., nuclear grading, presence and nature of necrosis, mitotic activity), whether they believed the features indicated a borderline diagnosis (between two diagnostic categories), and whether the pathologist would seek a consultative second opinion if they encountered the image during ordinary clinical practice.
To collect diagnostic information from participants after reviewing each image, we developed a histology form on the Qualtrics web-based platform; histology forms are commonly used by pathologists to record their observations and interpretations of images [31 (link), 41 (link)]. The histology form asked participating pathologists to provide the single most advanced diagnosis by selecting among the five diagnostic categories detailed above (ranging from benign to invasive carcinoma). They also rated the difficulty of interpreting the image and their confidence with their determination (both on scales from 1–7). Of less relevance to this paper, information was also gathered about histopathological features (e.g., nuclear grading, presence and nature of necrosis, mitotic activity), whether they believed the features indicated a borderline diagnosis (between two diagnostic categories), and whether the pathologist would seek a consultative second opinion if they encountered the image during ordinary clinical practice.
Free full text: Click here
