SUBJECTIVE BREAST-CANCER GRADING - ANALYSES OF REPRODUCIBILITY AFTER APPLICATION OF BAYESIAN BELIEF NETWORKS

OBJECTIVE: To examine the influence of Bayesian belief networks (BBNs) on the reproducibility of subjective breast cancer grading. STUDY DES IGN: Twenty samples were analyzed for intraobserver and 128 samples fo r interobserver reproducibility using the Bloom-Richardson and Helpap grading systems. The expression of diagnostic features was evaluated s ubjectively, and for each a decision it was determined to what extent it represented one of the different outcomes. Evidence was then entere d,for each diagnostic feature, into four different BBNs, recently desc ribed for breast cancer grading in the form of a relative likelihood r atio vector. RESULTS: With all cases considered, the use of decision s upport based on the Bloom-Richardson and Helpap grading systems did no t improve intraobserver reproducibility. This was found to be 68% and 80% in subjective gradings, respectively, and 60% and 70% in the BBN-s upported method. Interobserver reproducibility was not improved (58% a nd 70% in subjective gradings and 51-59% based on assessment with deci sion support). However, when only cases associated with high beliefs w ere considered, both intraobserver reproducibility (agreement rose fro m 68% to 93%) and interobserver reproducibility (agreement rose from 6 0% to 87%) of BBN-supported gradings exceeded the results of subjectiv e assessments. CONCLUSION: The results showed that the observers did n ot reach the same diagnosis (or grade) and that their observational as sessment of histologic features lacked agreement. Since BBNs reflected only the data entered, poor agreement existed in the contribution to the final diagnostic belief by the different features and, ultimately, in belief in the final decision.