Brought to you by EarthWeb
IT Library Logo

Click Here!
Click Here!

Search the site:
 
EXPERT SEARCH -----
Programming Languages
Databases
Security
Web Services
Network Services
Middleware
Components
Operating Systems
User Interfaces
Groupware & Collaboration
Content Management
Productivity Applications
Hardware
Fun & Games

EarthWeb Direct EarthWeb Direct Fatbrain Auctions Support Source Answers

EarthWeb sites
Crossnodes
Datamation
Developer.com
DICE
EarthWeb.com
EarthWeb Direct
ERP Hub
Gamelan
GoCertify.com
HTMLGoodies
Intranet Journal
IT Knowledge
IT Library
JavaGoodies
JARS
JavaScripts.com
open source IT
RoadCoders
Y2K Info

Previous Table of Contents Next


4.1. SMART SYSTEM

CBR has been most successful in the area of help desks, the most notable being SMART (Support Management Automated Reasoning Technology) (Acorn, 1992) for Compaq Computer Corporation. Compaq manufactures personal computer systems and one of its strategic business objectives is to provide quality customer support. Its Customer Support Center provides technical support, ranging from product information requests to problem resolution, on its wide variety of products. The workflow for the SMART System is shown in Figure 2.

When a customer calls in with a problem (e.g., printouts from the printer are faded), the details are gathered and entered into the SMART system. An initial search is then performed on the case library to locate previous cases with similar problem characteristics. Where information is incomplete, the system suggests further questions to be asked from the caller in order to determine more precisely, the category of the problem. Once a predefined threshold for matching is reached (i.e., the retrieved cases matches the new case by a certain percentage, say, 80%), the resolution from the retrieved case is then recommended to the caller. Hence, the SMART system makes use of previous cases to resolve new ones, instead of trying to diagnose a new problem from scratch. This has greatly enhanced the customer service and efficiency of the Customer Support Center by providing timely and accurate information to Compaq's customers within minimal response time. In addition, the system may be used as a training tool, while empowering the Support Center employees with more in-depth technical knowledge, thus increasing productivity.


FIGURE 2 Workflow of the SMART System.

Compaq has since extended the SMART system one step further by delivering intelligence and human expertise directly to the customers to help them solve their own problems, through the system known as QUICKSOURCE (Nguyen,1993). QUICKSOURCE is an example of knowledge publishing, in that the customer support function is put directly in the hands of the consumers. The system is sufficiently friendly to allow the customers to troubleshoot and solve their own problems, calling the Customer Support Center only as a last resort.

4.2. THE CARES SYSTEM

The Institute of Systems Science at the National University of Singapore is currently working on a research project with the Department of Colorectal Surgery at the Singapore General Hospital (SGH) to develop a system named Cancer Recurrence Support (CARES) system (Ong, 1997) that will predict the recurrence of colorectal cancer, using CBR as the primary technology. The CARES system employs CBR to compare and contrast between the new and past colorectal cancer patient cases, and makes inferences based on those comparisons to determine the high-risk patient groups.

The primary modality of treatment for colorectal cancer is surgery. However, although over two thirds of patients with primary disease undergo potentially curative surgery where all gross tumor is removed, up to 50% of these will eventually die in the ensuing 5 years, the majority from local, regional, or distant tumor recurrence. Adding to the problem is the difficulty in predicting the site of recurrence. This is, at the moment, difficult to do since primary colorectal cancers at different locations in the bowel may have different recurrence patterns.

The key to a successful follow-up program lies in the selective application of intensive follow-up and appropriate diagnostic tests and intensive chemotherapy for patients at high risk of cancer recurrence. Early identification of recurrence increases the effectiveness of therapy and survival of patients. Although the value of early detection of recurrence is recognized, the means by which this can be achieved is still controversial. It is argued that some form of follow-up is indicated for the remainder of the patient's life but the exact protocol to attain maximum benefit is debatable. In addition, many investigative tools, such as tumor markers, colonoscopy, imaging studies, and even exploratory surgery have added to the sensitivity of detection of recurrences, but if these were indiscriminately applied, it could lead to additional costs and patient morbidity without eventual survival benefit. Various statistical techniques have been proposed to identify high-risk patients; for example, a multivariate analysis of survival time using Cox's proportional hazard model to identify important prognostic factors. The CARES project, however, utilizes CBR to predict colorectal cancer recurrence by matching against past cases.

SGH currently has a clinical database (maintained on the hospital's mainframe computer) of more than 10,000 medical records of patients seen by the Department of Colorectal Surgery. The data acquired for each patient includes all relevant clinical and pathological information at the time of diagnosis and surgery as well as follow-up information that is collected according to a strict protocol. The historical data stored in this database is an invaluable resource that may contain unknown patterns, trends, or hidden meanings that can allow one to make predictions or forecasts. This hidden knowledge is not easily uncovered through conventional database queries. CBR is used in this application to match and retrieve past (i.e., existing) cases by recognizing their similarity to a new case. It makes assumptions about the present by looking at the past: "What was true yesterday is likely to be true today" (Kolodner, 1993). CBR is used to tap into the colorectal database to predict the risk of recurrence for a particular patient, based on matching his "profile" with the database records of previous patients who have known outcome (recurrence or no recurrence).

Most current medical Expert Systems are based upon expert knowledge and heuristics encoded directly as rules of the form:

"IF patient age > 60 yrs and has low fiber diet, THEN recurrence is likely"

For this application, namely the prediction of colorectal cancer recurrence, there is no clear body of heuristic knowledge available from which to build a rule set. Consequently, the use of CBR, which does not require knowledge to be made explicit in the form of rules, has clear advantages. The CARES system employs CBR to compare and contrast new and past colorectal cancer patient cases, and it then makes inferences based on those comparisons to determine high-risk groups.

The objective of the CARES system is to predict the length of disease-free period (number of months of no recurrence) after the first operation. In addition, the system is to suggest a follow-up regimen (tests) most appropriate for a particular patient. The objective here is to minimize the number of costly follow-up tests performed on a patient by customizing the test selection for that patient. To do this, the system will attempt to categorize the existing patients into groups according to which follow-up tests are most useful for recurrence prediction. A new patient can then be matched against these groups to recommend tests most appropriate or, alternatively, most inappropriate. The computation to support the recommendation will be done external to the CBR component. It should be noted that this functionality is highly dependent on the quality of data available; that is, there must be enough information already captured in the colorectal database such that the "categorization" of patients is possible. This project is currently on-going, and the estimated date of the implementation is October 1997.


Previous Table of Contents Next

footer nav
Use of this site is subject certain Terms & Conditions.
Copyright (c) 1996-1999 EarthWeb, Inc.. All rights reserved. Reproduction in whole or in part in any form or medium without express written permission of EarthWeb is prohibited. Please read our privacy policy for details.