The primary aim is to develop a protocol for enhancing the accuracy of image-based diagnostic systems, in radiology and other medical fields such as cytology and histology, by locating and eliminating sources of human error. The protocol will be presented in a document intended for publication as a book in the open literature. In addition to writing the protocol, the plan of work calls for conducting four illustrative studies to refine and illustrate the basic methods of the protocol. The first two of these studies focus on the problem of identifying those visual features of the medical image that are diagnostically informative and on finding ways to enhance the reader's ability to assess and merge those features into more accurate diagnostic judgments. The second and third studies illustrate gains in accuracy from repeated reading of the image, as made by the same and different readers. The fourth study is concerned with improving diagnostic accuracy through improved communication from the image reader to the referring physican. This request for a one-year continuation of the grant is necessitated by several complications encountered in conducting the illustrative studies. Considerably more time and resources have been required than originally anticipated for assembling adequately confirmed cases. Also, the feature space of Xeromammography, the modality chosen for the illustrative studies, turned out to be considerably more complex than originally anticipated. As a result, we have spent more time and resources than originally planned both in surveying and analyzing the features and in developing aids for enhanced image interpretation. Near-term objectives are to acquaint the medical community with systematic procedures for reducing error in image-based diagnosis, and to produce a practical guide for their implementation. The long-term objective is significant, cost-effective enhancement of diagnostic accuracy over a wide range of diagnostic settings. The complications we have encountered have increased the time and cost of the proposed work but do not otherwise present an impediment to reaching those objectives.
| Metz, C E; Shen, J H (1992) Gains in accuracy from replicated readings of diagnostic images: prediction and assessment in terms of ROC analysis. Med Decis Making 12:60-75 |
| Getty, D J; Pickett, R M; D'Orsi, C J et al. (1988) Enhanced interpretation of diagnostic images. Invest Radiol 23:240-52 |
| Swets, J A (1986) Form of empirical ROCs in discrimination and diagnostic tasks: implications for theory and measurement of performance. Psychol Bull 99:181-98 |
| Swets, J A; Feehrer, C E; Greenes, R A et al. (1986) Use of probability estimates in medical communications and decisions. Methods Inf Med 25:35-42 |
| Swets, J A (1986) Indices of discrimination or diagnostic accuracy: their ROCs and implied models. Psychol Bull 99:100-17 |
