Use of document image processing in cancer registration: how and why?

The aims of the study are to test and evaluate a system for document image processing (DIP) in cancer registration and coding. The results from use of optical character recognition (OCR) and image character recognition (ICR) in the data entry process are of particular interest. Cases of cancer must be reported both by clinicians and pathologists. Annually the Registry receives 80,000 reports concerning new cases and supplementary information on patients who are registered earlier [1]. Clinicians report new cancer cases to the Cancer Registry on a special designed form. Optionally, they can use a software-based application delivered from the Registry free of charge. As a part of the DIP-system, an optical digital image processing text (ODIT) system is used in recognition of machine- and hand-printed characters. The traditional registration and coding system is run parallel to the new system. After a period of testing the DIP-system nationwide, the new system will be evaluated and compared to the traditional system. The first part of the study compared the results from three character readers (CGK, XDR, and Nestor), which have been tested on our own application-specific data in a realistic setting [2]. The test shows that over 90% of the digits and about 70% of the letters can be correctly recognized by the system [3]. Feedback is communicated to clinicians to improve the quality of hand-print in the forms. It is to be hoped that this action among others will give a better recognition result for the next part of the study. The XDR Network Reader will serve as a reader in the ODIT-system chosen for the rest of the study. The software system Open Image Link shall serve as the image manager. De facto standard software like MS-Access is used as a part of the DIP-system. The next part of the study is the main project and started at the end of June 1994. Results from the project will be presented. Does the new system produce the expected high percentage of the clinicians will optionally use the software application instead of of paper forms in reporting cancer? And at what quality of data? Will the new DIP-system be cost-effective? The results from the study and the reasons for running such a project will be discussed.