Quality Checking and Mining Nephrology Biopsy Data

Abstract

One of the grave problems after kidney
transplantation is the recipient body’s immune
rejection of the transplanted organ. T -cells and
antibodies attack the organ, which in the worst case can
lead to graft failure. Biopsy is an important diagnostic
tool to evaluate a rejection episode. TBase© includes a
biopsy protocol that can easily be filled out by the
physician, which helped to collect and store 1447 biopsy
cases in the TBase© database. With respect to different
kinds of rejections, there exist some basic rules for the
entered data that are enforced during completion of the
protocol. Nevertheless, because so much biopsy data
was entered in by hand, it was necessary to check the
quality and plausibility of existing data with respect to
more complex rules that were not enforced during
protocol completion. In this paper, we present the
process of checking the biopsy data for consistency with
complex rules provided by an expert, as well as mining
of new rules using interpretable rule-based
classification methods. We discovered interesting rules
and relationships between features with respect to Tcell
mediated rejection (TCMR), antibody-mediated
rejection (AMR), interstitial fibrosis and tubular
atrophy (IF/TA), and polyoma virus (BKV)
nephropathy, with negative results concerning acute
tubular necrosis (ATN). The discovered rules further
support the quality and plausibility of the data, and
open avenues for further research.