Presented at the 1999 APIII Conference Return to 1999 Abstract Index
SURGICAL PATHOLOGY IMAGE-BASED DOCUMENTATION: REDUNDANCY VS. ESSENTIAL INFORMATION
Institute
of Molecular Pathology and Immunology
University of Porto
Porto, PORTUGAL
Domingos Cruz, MS
Background: In Surgical Pathology images
are essential in all steps of the process. The image selection
depends on several parameters related with the know-how
of who makes the acquisition, the specimen in analysis,
the morphological aspects and the stains used, among others.
On the other hand all the resultants image sets could be object of search and retrieval operations. The performance of retrieval is related with the amount and quality of information stored in the database.
We study the dependency of image-based documentation on parameters such as expertise levels of the users (acquisition, use for quality control), stains and sampled specimens. We also study the amount of information involved in image documenting of Surgical Pathology reports.
Design: With our system, five pathologists made the documentation of 1013 cases with the acquisition up to 12 digital images by case. The images acquired were stored along with the complete case data.
The image documentation was done in a way that image evidences could support diagnoses. Cases were randomly selected and reviewed by one or more pathologists. The level of agreement was registered. In absence of agreement the slides were reviewed using conventional light microscopy.
Results: We documented 1013 cases using 2809 images (2.7 images per case) occupying 863 MB of disk space. A total of 112 cases were selected for review based upon the related images (n=298), clinical and final diagnoses. The pathologists considered 92% of the images acceptable. In 11% of the cases the agreement with the original diagnoses was not possible and the slides were reviewed at microscope. The main problem to establish a correct diagnosis based on images was the incomplete documentation. In 25% of reviewed cases was founded 12.4% of redundant images.
Conclusion: The results showed that duplicated information in images (redundancy) and over-information within an image affect significantly the amount of information stored in Surgical Pathology databases. We did not find any relevant dependency between quantity of image-based information and topography or pathology. The results also showed the importance of the expertise level of the user responsible for acquisitions when using the database in Quality Control Assessment.
