Automated Annotation of Histopathology Data from Final Pathology Reports to Biorepository Samples
E Lickerman, J Sebastian, R Bowman, C Magyar, S Santoro, N Afsheen, A Sharif, S Dry. UCLA, Los Angeles, CA; Daedalus Software Inc., Cambridge, MA
Background: Rich histopathologic annotation enhances the value of tissue biorepository samples. Unfortunately, no automated, accurate annotation processes exist. Manual data entry underlies most current approaches. To solve this deficiency, we developed an automated procedure that extracts data from standardized sections of the finalized pathology report (PowerPath) and downloads the information to our biorepository database (Biomaterials Tracking and Management Application (BTMTM), Daedalus Software Inc.).
Design: Customized synoptic reports (SynR), based on College of American Pathologists recommendations, were developed for each gastrointestinal (GI) system organ. SynR are imported into the microscopic description section. Standardized language includes tumor types, grades and staging information. Non-productive text is deleted. Free text is limited to one field. The microscopic section text is imported from PowerPath into BTM. Then, synoptic report text is extracted, parsed and converted into specimen annotations. These annotations are entered into the Lucene text search engine as indices pointing to the surgical pathology case number. Each of these annotations can be used to search for samples from the relevant surgical pathology case.
Results: GI SynR, introduced in 2007, are used routinely by all UCLA GI pathologists. We permit personalized language in the final diagnosis section as SynR in the microscopic description section ensure standardized language. Oncologists and surgeons enthusiastically accepted SynR. Using pancreas as the test case, histopathologic data was imported successfully from the final PowerPath report into BTM and correctly associated with tissue samples. Samples with detailed histopathologic criteria are retrieved easily; for example, mixed acinar-endocrine + uncinate process + pT2. Complex searches using both SynR and other data fields elsewhere in BTM (ie, patient sex, specific research project subjects, etc) also are possible.
Conclusions: Our novel automated process uses SynR to extract standardized histopathology data from final PowerPath reports, then imports and links this data to relevant samples within our BTM biorepository database. Such systematic annotation greatly facilitates biorepository searches and eliminates known problems with free text searching. SynR are customizable by organ. GI pathologists and clinicians have accepted the SynR. Expansion to additional organ systems is being undertaken.
Monday, March 22, 2010 1:00 PM
Poster Session II # 261, Monday Afternoon