Evaluation of a decision support system for interpretation of myocardial perfusion gated SPECT
We have recently presented a decision support system for interpreting myocardial perfusion scintigraphy (MPS). In this study, we wanted to evaluate the system in a separate hospital from where it was trained and to compare it with a quantification software package.
A completely automated method based on neural networks was trained for the interpretation of MPS regarding myocardial ischaemia and infarction using 418 MPS from one hospital. Features from each examination describing rest and stress perfusion, regional and global function were used as inputs to different neural networks. After the training session, the system was evaluated using 532 MPS from another hospital. The test images were also processed with the quantification software package Emory Cardiac Toolbox (ECTb). The images were interpreted by experienced clinicians at both the training and the test hospital, regarding the presence or absence of myocardial ischaemia and/or infarction and these interpretations were used as gold standard.
The neural network showed a sensitivity of 90% and a specificity of 85% for myocardial ischaemia. The specificity for the ECTb was 46% (p < 0.001), measured at the same sensitivity. The neural network sensitivity for myocardial infarction was 89% and the specificity 96%. The corresponding specificity for the ECTb was 54% (p < 0.001).
A decision support system based on neural networks presents interpretations more similar to experienced clinicians compared to a conventional automated quantification software package. This study shows the feasibility of disseminating the expertise of experienced clinicians to less experienced physicians by the use of neural networks.
KeywordsImage interpretation Computer assisted Neural networks (computer) Radionuclide imaging Heart function tests Heart disease
- 1.Hachamovitch R, Berman DS, Shaw LJ, Kiat H, Cohen I, Cabico JA, et al. Incremental prognostic value of myocardial perfusion single photon emission computed tomography for the prediction of cardiac death: differential stratification for risk of cardiac death and myocardial infarction. Circulation. 1998;97:535–43. Erratum in: Circulation 1998;98:190.PubMedGoogle Scholar
- 3.Berman DS, Kang X, Van Train KF, Lewin HC, Cohen I, Areeda J, et al. Comparative prognostic value of automatic quantitative analysis versus semiquantitative visual analysis of exercise myocardial perfusion single-photon emission computed tomography. J Am Coll Cardiol. 1998;32:1987–95.PubMedCrossRefGoogle Scholar
- 11.Rumelhart DE, McClelland JL, eds. Parallel distributed processing, vols. 1 and 2. Cambridge: MIT Press; 1986.Google Scholar
- 13.Hanson SJ, Pratt LY. Comparing biases for minimal network construction with backpropagation. In: Touretzky DS, editor. Advances in neural information processing systems. San Mateo: Morgan Kaufmann; 1998. p. 177–85.Google Scholar
- 15.Riffenburgh RH. Statistics in medicine. London: Academic; 1999.Google Scholar