Related Experiment Videos
Accuracy of interpretation of microscopic images of cytologic, hematologic, and histologic specimens using a
K W Prasse1, E A Mahaffey, J R Duncan
1Department of Veterinary Pathology, College of Veterinary Medicine, University of Georgia, Athens, USA.
Objective:
To determine the accuracy of interpretation of microscopic images for pathologic study transmitted over Switched-56 lines using a desktop interactive video conferencing system.
Materials And Methods:
In subjective studies, two systems were connected using null-modem cables, which allowed evaluation of different bandwidths from 56 kbps to 384 kbps. Objective studies were done with two systems connected at distant sites via paired Switched-56 lines that produced an effective bandwidth of 112 kbps. A video camera mounted on a microscope was attached to the sending system. The resolution of the video image on the video conferencing system was 352 x 288 lines. Cases for cytology, hematology, and histopathology studies were selected from archives; one pathologist transmitted microscopic images, and a second pathologist made interpretations. The three pathologists were Board certified with similar experience that ranged from 20 to 35 years. Categories of interpretations or observations were predetermined for each study to allow the data on agreement between the direct microscopic interpretation or observation and that recorded by the receiving pathologist to be corrected for agreement attributable to chance alone. The results were analyzed using the kappa statistic.
Results:
In the subjective studies, image degradation prevented interpretation while the microscope stage was moved. This problem occurred at all bandwidths tested. Image quality limited microscopic details. Organisms < 1 micron in diameter could not be seen reliably. In objective cytologic studies, overall agreement was recorded on 89 of 99 observations. In the four categories of specimens, observed agreement ranged from 0.778 to 0.958, and kappa was 0.704 to 0.948. For hematology specimens, overall agreement was found on 69 of 80 observations; observed agreement on eight types of nucleated blood cells ranged from 0.5 to 1.0, and kappa was 0.429 to 1.0. Poorer color definition and image quality prevented accurate identification of lymphoblasts and eosinophils in particular. For histologic specimens, overall agreement was obtained on 56 of 66 observations, observed agreement on four categories of histologic change ranged from 0.73 to 0.93, and kappa was 0.47 to 0.9.
Conclusions:
The desktop interactive video conferencing system, as configured in this study, was unsuitable for making definitive diagnoses from transmitted microscopic images.