|
||||
Home
» Archive
» 2023
TDK conference 2023László Fruzsina - year 6 University of Veterinary Medicine Budapest, Department of Pathology Supervisor: Dávid Géza Horváth DVM The mastocytoma or mast cell tumor is the most prevalent skin tumor in dogs, most frequently affecting the cutis or the subcutis. The clinical stage is one of the most relevant prognostic factors which has a strong association with the hystological tumor grade. Hence it is vital to determine the tumor grade to adequatly plan the therapy and the surgical care. Among the examination aspects of the grade, the mitotic cell count can be considered the most informative, because of its high prognostic value. To define this, digital pathological systems has been made available. During our examination we have compared the performance of the QuPath software's and a pathologist's manual grading by digitally identifying the mitotic cell forms. In 19 of the 39 cases (48.7%) the manual classification was consistent with the digital grading. The Cohen’s kappa was -0.284, hence there was no agreement between the algorithm and the manual observer. The algorithm training was not done by an expert, therefore the discrepancy was probably due to the lack of dataset with sufficient quality for the program. The implementation of artificial intelligence and digital pathology systems in diagnostics undoubtedly ease the work of pathologists, nevertheless as it is aslo shown by our results human professionals cannot be ignored. Most probably an algorithm trained by an experienced pathologist will be able to provide a more efficient performance in diagnostics in the foreseeable future, when the pathologists can focus on the validation of the diagnosis. List of lectures |