Probabilistic Diagnosis Using a Reformulation of the INTERNIST-1/QMR Knowledge Base

Abstract

We have developed a probabilistic reformulation of the Quick Medical Reference (QMR) system. In Part I of this two-part series, we described a two-level, multiply connected belief-network representation of the QMR knowledge base and a simulation algorithm to perform probabilistic inference on the reformulated knowledge base. In Part II of this series, we report on an evaluation of the probabilistic QMR, in which we compare the performance of QMR to that of our probabilistic system on cases abstracted from continuing medical education materials from Scientific American Medicine. In addition, we analyze empirically several components of the probabilistic model and simulation algorithm.

Key-Words

Evaluation - Expert Systems - Computer-Aided Diagnosis - Probabilistic Inference - Belief Networks

^* QMR is a registered trademark of the University of Pittsburgh.

^® I We are currently using the INTERNIST-I KB (circa 1986). rather than the more recent QMR KB. These two KBs are quite similar. to the extent that the methods in this paper can be applied to the latter KB as well. For simplicity. where the distinction between the INTERNIST-J KB and QMR KB is inconsequential. we will refer to the INTERNIST-I KB as the QMR KB.

• REFERENCES

• 1 Cooper GF. The computational complexity of probabilistic inference using Bayesian belief networks. Artif Intell 1990; 42: 393-405.
• 2 Rubenstein E. Personal communication. 1990
• 3 Miller RA. Personal communication. 1989
• 4 Miller RA. Personal communication. 1990
• 5 Miller RA, Masarie Jr FE. The demise of the “Greek Oracle” model for medical diagnostic systems. Meth Inform Med 1990; 29: 1-2.
• 6 Miller RA, Pople HEJ, Myers JD. Internist-1: An experimental computer-based diagnostic consultant for general internal medicine. N Engl J Med 1982; 307: 468-76.
• 7 Miller R, Masarie FE, Myers JD. Quick medical reference (QMR) for diagnostic assistance. MD Computing 1986; 03: 34-48.
• 8 Ott L. An Introduction to Statistical Methods and Data Analysis. Boston, MA: PWS-Kent Publishing Company; 1988
• 9 Wyatt J, Spiegelhalter DJ. Evaluating medical expert systems: what to test and how?. Med Inform 1990; 15: 205-17.
• 10 Bankowitz RA. User variability in abstracting and entering printed case histories with QUICK MEDICAL REFERENCE (QMR). In: Proceedings of the Eleventh Annual Symposium on Computer Applications in Medical Care. Stead WW: (ed) Los Alamitos, CA: IEEE Comp Soc Press; 1987: 68-73.
• 11 Henrion M. Towards efficient probabilistic diagnosis in multiply connected networks. In: Influence Diagrams, Belief Nets and Decision Analysis. Oliver RM, Smith JQ. (eds) Chichester: Wiley; 1990: 385-407.
• 12 Cooper GF. Bayesian Belief-Network Inference Using Recursive Decomposition. Knowledge Systems Laboratory Memo no. KSL-90-05. Stanford, CA: Stanford University; 1990
• 13 Pearl J. Evidential reasoning using stochastic simulation of causal models. Artif Intell 1987; 32: 245-57.
• 14 Shachter RD, Peot M. Simulation approaches to general probabilistic inference on belief networks. In: Machine Intelligence and Pattern Recognition: Uncertainty in Artificial Intelligence 5. Henrion M, Shachter R, Kanal LN, Lemmer JF. (eds) Amsterdam: North-Holland Publ Comp; 1990: 221-31.
• 15 Shwe MA, Cooper GF. An empirical analysis of likelihood-weighting simulation on a large, multiple connected medical belief network. Comp Biomed Res. 1991 to appear.
• 16 Bankowitz RA, McNeil MA, Challinor SM, Parker RC, Kapoor WN, Miller RA. A computer-assisted medical diagnostic consultation service: Implementation and prospective evaluation of a prototype. Ann Intern Med 1989; 110: 824-32.
• 17 Miller RA, McNeil MA, Challinor SM, Masarie FEJ, Myers JD. The INTERNIST-1/QUICK MEDICAL REFERENCE project - Status report. Western J Med 1986; 145: 816-22.
• 18 Heckerman DE, Horvitz EJ, Middleton B. An Approximate Nonmyopic computation for Value of Information. Knowledge Systems Laboratory Memo no. KSL-91-15. Stanford, CA: Stanford University; 1991