ISSN: 2278 – 1323                                      International Journal of Advanced Research in Computer Engineering ...
ISSN: 2278 – 1323                                      International Journal of Advanced Research in Computer Engineering ...
ISSN: 2278 – 1323                                      International Journal of Advanced Research in Computer Engineering ...
ISSN: 2278 – 1323                                      International Journal of Advanced Research in Computer Engineering ...
ISSN: 2278 – 1323                                     International Journal of Advanced Research in Computer Engineering &...
ISSN: 2278 – 1323                                   International Journal of Advanced Research in Computer Engineering & T...
Upcoming SlideShare
Loading in …5
×

382 387

238 views

Published on

Published in: Health & Medicine, Business
0 Comments
0 Likes
Statistics
Notes
  • Be the first to comment

  • Be the first to like this

No Downloads
Views
Total views
238
On SlideShare
0
From Embeds
0
Number of Embeds
2
Actions
Shares
0
Downloads
2
Comments
0
Likes
0
Embeds 0
No embeds

No notes for slide

382 387

  1. 1. ISSN: 2278 – 1323 International Journal of Advanced Research in Computer Engineering & Technology Volume 1, Issue 4, June 2012 AN ENHANCEMENT IN RELEVANCE KNOWLEDGE DISCOVERY MODEL FOR MEDICAL REASONING USING CBRM Prof.S.Anithaa anithabennett@yahoo.co.in VIT university Chennai campus the need for specific diagnostic tests in specificAbstract: Healthcare technology is entering in to a patients, leading to effective dispensing of health carenew evolutionary phase. The medical community has measures and eliminating unnecessary tests whichan obligation to the public to provide the safest, most leads to saving valuable time and cost considerably.effective possible healthcare. In this contextKnowledge sharing is crucial for better patient care in Keywords: Case Based Reasoning Model, Healththe healthcare industry, but it is challenging for care, Knowledge based systems, Knowledge sharing.physicians to exchange their clinical insights,outcomes and valuable experiences, particularly with I. INTRODUCTIONregard to make decisions in diagnosing disease atearly stages. The aim of our study is to facilitate Medical diagnosis and decision-making involvesknowledge sharing and information exchange in this interplay between vast numbers of medical knowledgearea by means of a knowledge-based system. We resources . This can range from implicit knowledgepropose a knowledge-based system, which held by healthcare workers to experiential and data-automatically models each physician’s experience and induced knowledge. Systems that can simultaneouslyexperts in health care industry. This is done by access and combine relevant information from thesecollecting as many as possible instances occurred various knowledge resources are crucial to theduring past histories about the patient symptoms, diagnostic and prognostic processes and subsequentlytreatments, and relevant medications. This process to the efficient treatment of patients. From a decisioncan be referred as Case Based Reasoning support viewpoint, healthcare workers need complete,Model(CBRM) and will be analyzed from a statistical contextually-relevant information that is consistentperspective to form a authenticated interactive with the patients current medical state and that isknowledge sharing process for making right decisions appropriately presented at the correct level ofat right time. abstraction. In this research we have drawn on medical knowledge management initiatives thatKey Application Areas includes : Improving the promote the collection, integration and distribution ofQuality of Patient Care Identifying high-risk patient a single medical modality [1]. This allows us to buildgroups with combinations of symptoms and/or risks, encapsulated patient profiles that are used both toIdentifying the need for prophylactic measures to effectively store patient data and also for the purposesprevent outbreak of disease, Improve patient care of comparison with new patient profiles for diagnosisthrough efficient prescribing of drugs by identifying and treatment. The system we are developing also hasduplication or over-prescribing of drugs, and also beneficial repercussions from a healthcare modelingidentifying potential drug interactions in view point [2], as explanatory models from amassedcontraindicated drugs, Search for statistical data patient data can easily be created that can identifyregarding patient-disease patterns, classifying them trends as well as comparing diagnosis, treatments andbased on age, gender, geographical locations, food departments.groups, etc., by identifying common factors amongpatients with similar diseases. Identifying the need for II. RESEARCH ISSUES AND PROPOSEDdiagnostic tests in specific patients, leading to CONTRIBUTIONeffective dispensing of health care measures. Clinical data mining has three objectives:Revenue Generation and Saving Time Lowering the understanding the clinical data, assist healthcarecost and effort involved in clinical Research and professionals, and develop a data analysisDevelopment through automated reviews. Identifying methodology suitable for medical diagnosis[3]. Our work deals with the insight of health care 382 All Rights Reserved © 2012 IJARCET
  2. 2. ISSN: 2278 – 1323 International Journal of Advanced Research in Computer Engineering & Technology Volume 1, Issue 4, June 2012professionals such as radiographer or physician worthy challenges and is driving CBR researchdecisions, as they diagnose and treat patients during forward by offering a variety of complex tasks, whichillness. Since the health care professionals interacts are difficult to solve with other methods andwith the patient profile (patient age, gender, medical approaches. [5] The origin of CBR can be traced tohistory, examination and laboratory findings, medical Yale University and the work of Schank and Abelsonimagery like X-rays etc are integrated as encapsulated in 1977 [6]. Earlywork exploiting CBR in the medicalprofiles) in the course of a diagnosis the system domain was performed by Koton [7] and Bareiss [8]analyses their actions. This enables to capture human in the late 1980s. The CBR is inspired by humanexpertise and proficiency in diagnosing and treating reasoning, i.e., solving a new problem by applyingany particular disease. This proficiency of diagnosing previous experiences adapted to the current situation.disease and relevant treatment is referred as an expert A case (an episodic experience) normally contains aknowledge and is stored in knowledge base for future problem, a solution, and its result. The CBR is anreference. This information can then be used to filter, appropriate method to explore in a medical contextretrieve and display the most relevant similar patient where symptoms represent the problem, and diagnosiscase histories from huge repositories of patient data and treatment represent the solution. With itsand can further be used for diagnostic comparison capability of incrementally collecting, reusing andwith new patient complaints, symptoms or other sharing the knowledge implicitly embedded inclinical evidences. The information of new patients previously experienced situations, case-basedcan be entered in to the system and previous reasoning (CBR) [9] is currently recognized as a verydiagnoses, treatments and outcomes for similar well suited reasoning methodology in medicalpatients can be instantly accessed by physicians. The applications.application can save much physician time by avoidingmanual scanning of patient records and retrieved IV. CBRM CASE BASED REASONING MODELinformation can be interactively explored and canused to guide health care professionals towards Life cycle of CBRM is a combinational model ofappropriate and relevant information regarding statistical view and CBR view. This combinationaldiagnoses and treatments. model gives knowledge for health care professionals to make decisions as shown in figure 1. The CBREffective use of this knowledge base of previous case view has four main steps (4R‟s) retrieve, reuse,histories is made possible by the application of Case- revise, and retain. In the retrieval step, a new problemBased Reasoning (CBR) techniques. CBR is a well is matched against the previous cases in the caseestablished method for building medical systems [4], library. Domain knowledge is used to determine theand one of the intuitively attractive features of CBR in similarity between the new case and the case availablemedicine is that the concepts of patient and disease in domain knowledge, and the degree of similaritylend themselves naturally to a case representation. leads to make decision about the newly arrived case.Also medical practitioners logically approach The most relevant solutions available in thediagnosis from a case-based standpoint (i.e., previous knowledge base are proposed to solve the newlyspecific patient interactions are as strong a factor as arrived problem with respect to some adaptations ifindividual symptoms in making a diagnosis). There necessary. The selected solution is revised before it isare three main advantages in our approach. First by reused. Then, the new problem and its solution arereusing collective knowledge in support of similar retained in the case library for future use.patient cases the time required to diagnose or treat anew patient can be significantly reduced. Second, the CBR systems typically needs to undergoapproach facilitates knowledge sharing (remote or preprocessing and filtering prior to Case formulation.otherwise) by retrieving potentially relevant For example, if the clinical data are collected fromknowledge from other experiences. Finally from a sensor signals, images, free-text sources, etc., then theknowledge management perspective, contextual system may require feature extraction, feature mining,expert knowledge relating to particular cases may now indexing, weighting, etcbe stored and reused as an additional resource forsupport, training and preserving knowledge assets. III. CASE BASED REASONINGCase Based Reasoning (CBR) is a recognized andwell-established method for the health science. Thehealth science domain offers the CBR community 383 All Rights Reserved © 2012 IJARCET
  3. 3. ISSN: 2278 – 1323 International Journal of Advanced Research in Computer Engineering & Technology Volume 1, Issue 4, June 2012 similarity relationship between cases are grouped into a set of independent attributes such as gender, age, marital status, blood pressure, depression, smoke, stress, anxiety etc. For each attributes, a relevance metric is defined to measure the similarity between two items. For example, two cases with same gender will get the maximum similarity rating while for attributes that greatly dissimilar, will get low rating. The output of each metric is an integer value. The degree of similarity is expressed between 0 (not similar) and 1 (very similar) . In this paper, we have created two sample cases for the similarity retrieval purpose in case database. Figure 5 shows the similarity comparison between New Case and Sample 1 while Figure 6 shows the similarity comparison between New Case and Sample 2. New Case Age 24 Sex Male Marital Status N Blood Pressure Y Depression Y Smoke Y Stress Y Anxiety YFigure 1:CBRM (Case Based Reasoning Model) Food Habit V Alcohol YIn the medical domain, the health care professionals,clinicians or doctors may start their practice with Figure 2: Summary of new problems entered bysome initial experiences (solved cases). Over a period Health care professionals and labeled as „New Case‟of time, they use that initial experiences as pastexperiences to solve a new problem. This may involve Sample 1some adjustment of the previous solutions to solve the Age 32new problem. Thus, a new experience (case) has been Sex Malecreated, which enriches the clinician‟s / doctor‟s set of Marital Status Yexperiences. In fact, this is how the traditional CBR Blood Pressure Ycycle works. So, the CBR is a reasoning process, Depression Ywhich is medically accepted and also getting Smoke Nincreasing attention from the medical domain. Stress Y Anxiety N V. RELEVANCE COMPUTATION Food Habit NV Alcohol YThe CBRM accepts new cases from health careprofessionals and preprocess them in order to remove Figure 3 shows the summary sample1noise, inconsistent data. The outcome ofpreprocessing is sent to the CBR model whichattempts to retrieve the most similar case from caselibrary. The system will carry out similaritycomputation test by weighted average using nearestneighbor algorithm. This algorithm will calculatesimilarity between new problem and similar cases incase library. The nearest case with highest k score willbe used as similar case to solve new problem. The 384 All Rights Reserved © 2012 IJARCET
  4. 4. ISSN: 2278 – 1323 International Journal of Advanced Research in Computer Engineering & Technology Volume 1, Issue 4, June 2012Sample 2 The Followings are similarity calculation (relevanceAge 45 computation) by weighted average using nearestSex Male neighbor algorithm for both sample cases (weight 5Marital Status Y for high importance and weight 1 for low importance):Blood Pressure YDepression Y Similarity computation: New Case and Sample 1Smoke Y = 1/10 * [(1 * 0.8) + (1 * 1.0) +(1*0.1)+ (5 *Stress Y 1.0)+(5*1.0)+(5*0.1)+(5*1.0)+(5*0.1)+(5*0.1)+(5*1.Anxiety N 0)]Food Habit V =1/10*(0.8+1.0+0.1+5.0+5.0+0.5+5.0+0.5+0.5+5.0)Alcohol N =1/10*(23.4)Figure 4 shows the summary sample2 = 2.34New Case VS Sample 1 Similarity computation: New Case and Sample 2 = 1/10 * [(1 * 0.7) + (1 *1. 0) +(1*0.1)+ (5 * 1.0)New Case Similarity Sample 1 +(5*1.0)+ (5*1.0) +(5*1.0)+ (5 * 0.1) + (5* rating 1.0)+(5*0.1)] = 1/10*(0.7+1.0 +0.1+5.0+5.0+5.0+5.0+ 0.5 +5.0 +Age 24 0.8 Age 32 0.5)Sex Ma 1.0 Sex Male =1/10*(27.8) le = 2.78Marital N 0.1 Marital YStatus Status The result of similarity computation will be used toBlood Y 1.0 Blood Y determine the adoption of similar case‟s solution asPressure Pressure solution to new problem. The acceptance level isDepression Y 1.0 Depressi Y predetermined by expert committee. From the on similarity calculation, the system will choose solutionSmoke Y 0.1 Smoke N from Sample 2 as solution for New Case (2.78 >Stress Y 1.0 Stress Y 2.34). The system will reuse previous solution toAnxiety Y 0.1 Anxiety N solve users‟ current requirement. However, if level ofFood Habit V 0.1 Food NV acceptance is rejected, the system will send current Habit problem case to the expert committee. After receivedAlcohol Y 1.0 Alcohol Y solution from committee, the system will makeFigure 5: shows New Case VS Sample 1 with proposal to users.. The proposed solution will besimilarity measures compared and revised against the actual result. If the result is accepted, the system will learn new case andNew Case Similarit Sample 2 retain the case in case library for future use. If the y evaluation is rejected, the system will close the case rating and users will be advised to make consultation withAge 24 0.7 Age 45 expert committee. Finally the solution was verifiedSex Mal 1.0 Sex Mal with the statistical methods and the outcome is e e updated in case library. At the heart of this model liesMarital N 0.1 Marital Y a knowledge discovery method that permits relevanceStatus Status knowledge to be automatically extracted from existingBlood Y 1.0 Blood Y structured and unstructured data that are available inPressure Pressure preprocessing stage. Not only this model can handleDepression Y 1.0 Depression Y data format diversity, high dimensionality, andSmoke Y 1.0 Smoke Y relative importance of the data source, but it is alsoStress Y 1.0 Stress Y capable of incrementally updating the existingAnxiety Y 0.1 Anxiety N indexing structure when new cases are added to the system. The techniques outlined in this paper willFood Habit V 1.0 Food Habit V yield significant benefit in the improvement ofAlcohol Y 0.1 Alcohol N decision support tasks and provide a better insight into the process of medical reasoningFigure 6: New Case VS Sample2 385 All Rights Reserved © 2012 IJARCET
  5. 5. ISSN: 2278 – 1323 International Journal of Advanced Research in Computer Engineering & Technology Volume 1, Issue 4, June 2012 VI. CHALLENGES OF CASE BASED committee for confirmation on the solution and REASONING verified with the statistical methods. Future study will be focused more on preprocessing of input data to theThere are a huge number of medical applications CBRM which results in further improvement of thewhich offers challenges for the CBR researchers and system.drive advances in research. Some of the importantresearch issues are given in the following. References [1] Jadad, A. R., Haynes, R.B., Hunt, D. & Browman,1)Feature extraction is becoming complicated in the G.P. "The Internet and Evidence-Based Decisionrecent medical CBR systems due to a complex data Making: A Needed Synergy for Efficient Knowledgeformat like sensor data, images, time series data , and Management in Health Care,," Canadian Medicaldata with free text format. Association Journal, Vol. 162(3):362-5, 2000.2) Feature selection and weighting are two other [2] Ivatts, S. & Millard, P.H. "Health care modelling -important factors for which many CBR systems Why should we try?," British Journal of Health Caredepend on expert knowledge. Cases with hidden Management, Vol. 8(6), pp.218-222, 2002.features could also affect the retrieval performance. [3] Clinical Data Mining: a Review J. Iavindrasana et3) The component that plays a central role in the CBR al, yearbook 2009.systems is the case data base or case library. A casebase can be considered as concrete knowledge of a [4] Nilsson, M. and Sollenborn, M., "Advancementsmodel consisting of specific cases. The cases stored in and Trends in Medical Case-Based Reasoning: Ana case library should be both representative and Overview of Systems and System Development," Incomprehensive, so as to cover a wide spectrum of Proceedings of the 17th International FLAIRSpossible situations Conference, pp.178-183, 2004.VII. CONCLUSION AND FUTURE WORK [5] Shahina Begum et al. IEEE TRANSACTIONS ON SYSTEMS, MAN, AND CYBERNETICS—The current system in hospitals whereby doctors enter PART C: APPLICATIONS AND REVIEWS, VOL.patient information using paper charts is cumbersome, 41, NO. 4, JULY 2011time-consuming and does not facilitate knowledgesharing. Different types of information, including [6] R. C. Schank and R. P. Abelson, Scripts, Plans,imagery, are stored in different locations and valuable Goals and Understanding. Erlbaum, Hillsdale, Newtime is often lost trying to correlate data in order to Jersey, 1977.diagnose and treat patients. This system can addresssuch issues by providing doctors with instant access to [7]P. Koton, “Using experience in learning andinformation that will allow them to make critical problem solving. Massachusetts Institute ofdecisions and prognoses with greater speed and Technology,” Ph.D. dissertation, Lab. Computerefficiency. It facilitates knowledge sharing and Science, Dept. Electr. Eng. Comut. Sci.,supports effective communication about the most MIT/LCS/TR-441, Cambridge, 1989.effective ways in which to treat patients by linkingsimilar patient case histories using case-based [8] R. Bareiss, “Examplar-based knowledgereasoning techniques.[10]. The concept of CBRM is a acquisition: A unified approach to concept,service oriented model which retains previously classification and learning,”Academic Presssolved cases and experts‟ knowledge in case library. Professional, Inc., San Diego, CA, 1989.The system can propose wellness solution that ispersonalized to users based on the adaptation of [9] Kolodner JL. Case-based reasoning. San Mateo,previous solution in similar cases. The proposed CA: Morgan Kaufmann; 1993.CBRM uses a combination of Case Based Reasoningand Statistical Models. Furthermore, the advantage of [10] David Wilson,et al, 3rd International IEEEthis model over the other model is, in other CBR Conference Intelligent Systems, September 2006models the limitation was that the system will not beable to propose solution to new case that has nosimilar solution in Case Library. But in our Model thesystem will build new case and send to the expert 386 All Rights Reserved © 2012 IJARCET
  6. 6. ISSN: 2278 – 1323 International Journal of Advanced Research in Computer Engineering & Technology Volume 1, Issue 4, June 2012Author:Prof.S AnithaaVIT UniversityChennai campusTo my credit I have published more than 12 papers inNational and International journals and conferences.My area of interest are Data mining, Databasemanagement systems, Knowledge Engineering 387 All Rights Reserved © 2012 IJARCET

×