KNOWLEDGE BASED RADIOLOGY INFORMATION SYSTEM

点击翻译按钮获取中文摘要

The goal of this project is to improve the accuracy of the radiologist as he/she interprets x-ray examinations. Two computer-based approaches to this goal will be studied. First a computer-generated, patient-specific consult will be made availabe to the radiologist at the time he reads the x-ray. It will consist of a brief clinical summary whose elements are chosen for their information content, a differential diagnostic list containing the patients 1 to 5 most likely diseases, and a list of prompts indicating those abnormal x-ray findings that would contribute the most to diagnosing the patient. The second technique investigated will be a computer-managed, continuous audit that will use the discharge diagnoses of each patient to help determine the average accuracy of the x-ray interpretations. This mechanism can provide feedback to the individual radiologists concerning the specific diseases and findings for which their readings are suboptimal. Evaluation will be through comparison of the chest x-ray interpretations generated with and without the consult/prompt system. Two approaches will be used. The first will compare all examinations from a 12-month period before implementation with those from a similar period after implementation. The second will assess the difference in interpretations on 200 select chest x- rays interpreted by six physicians with and without the consult and prompts. For this test, an expert panel will determine the actual abnormalities present. The collected interpretations of the 200 films will be used to test the audit system to determine whether its results conform with the results of traditional measures of performance based on rereading multiple x-rays. The applications created during this project will use the expert system tools in the HELP system combined with a group of knowledge management routines. An implicit goal of the project is to investigate the application of information theory to the problem of accuracy in the radiology department. These tools will have general applicability in medicine.