Computer-aided clinical laboratory diagnosis in conjunction with the electronic medical textbook.

1. INTRODUCTION. Medical knowledge has been increasing and diversifying on a worldwide scale, while the specialization of physicians has been extended vigorously. Under this environment, it may be natural that mistakes are made in a comprehensive diagnosis, as the physician cannot master all of this dramatically increasing volume of knowledge. The knowledge has extended beyond the memory of human beings, thereby causing the deterioration of service; this is called the "Knowledge crisis." To tackle this problem, the Electronic Medical Textbook (EMT) has been conceived and set up as a medical knowledge base for physicians to optimize both their specialties and activities in clinical practice. Meanwhile, laboratory information systems were widely introduced. However, there are few systems which allow interpretation of the findings obtained. With this in mind, we have improved the utility of the EMT by enhancing its function with laboratory information follow-up, thesaurus back-up, Japanese language support, and online access. 2. SYSTEM DESCRIPTION. The knowledge database for medical decision-making consists of three categories: 1) Medical domain knowledge (including approximately 3500 diseases) from the AMA's "The Current Medical Information & Terminology (CMIT)"; 2) Knowledge on relationship between laboratory testing results and diseases from "The Effects of Disease on Clinical Laboratory," compiled by the AACC; and 3) Clinical testing knowledge from Otsuka's laboratory test handbook "Kensa-Kojien." These categories are connected by links in the process of cross-reference. In actual use, the first is to select the supporting system bringing up clinical signs and findings on CRT from which users can then choose any representations corresponding to the patient's clinical state. Once the relevant objects have been selected, the system presents the correlated investigative tests to be performed, along with a scope of laboratory tests ordering for its initial investigative task. The required tests against the data objects denoted by "High," "Low," or "Abnormal" are linked to possible diagnoses, which appear on CRT in order of likelihood. Similarly, the possible diagnoses can be obtained directly by consulting the patient's laboratory test results in linkage, by way of automated data transformation into the corresponding data object using the reference interval. If necessary, additional information relative to clinical signs can be added. With the repetition of this procedure, clinically useful tests can be located, thus increasing the likelihood of an appropriate diagnosis. The proposed diseases can be confirmed through cross-reference to the patient's clinical signs from the description in the Textbook on CRT. In order to bolster the efficiency of cross-checking on CRT, the sentences including hit-words are highlighted in red. Moreover, with regard to specific keywords for disease, those sentences are highlighted in green to emphasize differences in between. In parallel, the description points out an active behavior of laboratory test results in the progression of the disease and is also able to display applicable laboratory test listings. Furthermore, this knowledge-based interpretation system is directly accessible via online network services in referring to laboratory test results. 3. CONCLUSION. A combination of the CMIT and laboratory diagnostic information in the knowledge database allows the narrowing down of proposed diagnoses due to its flexible approach, which depends not only on clinical signs and findings, but also on applicable test-ordering support and suggestions. This enhances the potential usefulness of the EMT. The system contributes to the support of the diagnostic procedures and assists in educational setting, because the process can be operated in reverse based on clinical signs and laboratory findings in a dialogue style on CRT. The related disease description highlighted in color has won a good reputation dur