Information Security of Patient E-Records Exchange
Jonathan J. R. Chen
Department of Information Management Van Nung Institute of Technology Jung-Li, Tao-Yuan, Taiwan, R.O.C.
Yuanchi Liu
Department of International Trade Ta-Hwa Institute of Technology Chiung-Lin, Hsin-Ju, Taiwan, R.O.C.
Horng-Min Tsai
Graduate School of Resource Management National Defense Management College
Jung-Ho, Taipei, Taiwan, R.O.C.
Abstract
The volume of traditional medical records increases linearly along with the increasing number of patients and gives a burden to medical costs. This problem can be solved by the introduction of e-records if information security for associated privacy can be warranted. Given the rapid progress of information technology, competition in the medical industry, and national health insurance, the research of information security for e-record is imperative. This paper uses group digital signature to provide an easy and friendly mechanism of information security for the privacy of both patients and doctors. Our mechanism is helpful to the information security of e-medical industry.
I. Introduction
Traditional medical records for patients need a huge volume of paper. The readability of doctor handwritings will fade away as time passes by. The volume of traditional medical records increases linearly along with the increasing number of patients and gives a burden to medical costs. This problem
can be solved by the introduction of e-records if information security for associated privacy can be warranted. In recent years, researches focus on the privacy of patients [1,9,11,12]. Record privacy involves both patients and doctors. In addition, how can the records be exchanged over Medical Information Exchange Center (MIEC)? To what extent has the mechanism for the associated information security been done? Literature discussions of the security [10,15] are very limited.
This paper proposes a mechanism for exchanging e-records over medical computer network.
II. Installation of e-records
The information center of Hospital A selects a big prime with 1,024 bits p≡4p1q1+1 where p1,q1 are big primes of 512 bits. Select modulo(p) with a primitive element over GF(p) with order
1 1q
p (6), select a number xA , and calculate
( p )
g
y
A≡
xAmod
. {p
,q
, yA } and { p1,q1,xA} are the public key of the system center and private key, respectively. Given the system parameters, doctorA
i in Hospital A selects a numberx
i , calculatesy
i≡ g
xi( mod p )
, and2 registers {
y
i} to the system center as the public key.The system center selects a number
k
i, calculates( p )
g
r
i≡
kimod
and( mod p
1q
1)
y k r x
s
i≡
A i+
i i , and sends {r
i,s
i} through a safety channel to doctorA
i where {x
i} is the private key of doctorA
i and {r
i ,s
i} is semi-private key. Note that semi- private key is defined as the public (private) keys set between doctorA
i and system center. No one else knows the semi-private key.III. E-Record Algorithm
Patient a sees doctor
A
i. Let record for a is m written by doctorA
i. The algorithm is as follows:(1) Select a number σ1 and calculate
( p )
g
r ′ ≡
σ1mod
( p )
r r
r ′′ ≡
i′ mod
( )
[ y r mod p ] ( mod p
1q
1)
y
i′ ≡
i′
(2) Select a number σ2 and calculate
( p )
g
r ′′′ ≡
σ2mod
( x
1) r
2s
2( mod p
1q
1)
m ≡
i+ σ ′′′ + σ
(3) Group digital signature for Doctor
A
i is {m
i,y′
i,r ′′
,r ′′′
,s1,s2}IV. The Algorithm for transferring e-records
Assume that the original record for patient a is kept in Hospital A. For some reasons, patient a goes to Hospital B and Hospital B needs his/her record.
The algorithm to transfer e-record is as follows:
(1) Patient a agrees to transfer his/her record from Hospital A to Hospital B by digital signature for medical reference. Let a1 be digital certificate.
(2) Given digital certificate a1 presented by patient a, digital certificate
B
∗ can be granted if doctorB
j , staff doctor, and director approve the transfer of e-record. The algorithm of group digital signature is as follows:(3) The system center or director of Hospital A
checks the validity of digital certificate
B
∗ and sends {m
i,y′
i,r ′′
,r ′′′
,s1,s2} to Hospital B after encryption, e.g., ElGamal [7], RSA [13].(4) The system center or director of Hospital B checks the equations
g y
Ar( ) ( r
yip )
s1
≡
′′′′
′mod
g
( ) ( ) (
yi r r s p)
m ≡ ′ ′′′ ′′′ 2 mod
If both equations are valid, accept e-record m. If not, reject m as shown in the diagram as follows:
V. Analysis of system information
security The information security can be analyzed as follows.(1) Installation of e -record
Given the fact that in factorization [13], the composite number of 512 bits has been defeated [3]. For 30 years in the future, the development of information technology should use the composite number of 1,024 bits and install the encryption system by the use of discrete logarithm [7] to prevent possible attacks.
Thus, our algorithm is helpful to patient privacy Director Waden
Hospital AA A A醫院
Hospital
MIEC
doctor
B
je-record e-record
Doctor
A
ie-record e-record
Hospital
Hospital
Hospital
Director
Staff doctor D
and medical information security.
(2) E-record algorithm
The security of group digital signature [2,4,5,8,14] amounts to installation of e-records.
Hospital
A
i uses specific digital signaturey′
i for each record m. Thus, doctors of other hospital, e.g., Hospital B cannot tell whethery′
i andy
i′′
are written by the same anonymous doctor.(3) The algorithm for e -record transfer
In this section, the ElGamal theory [7] is used to generate digital certificate a1 for patient a.
Let multiple digital signatures in the same hospital be
B
∗ generated from MIEC.Information security is based on the installation of e-records. Thus, the encryption/ decryption between hospitals should be free from attacks.
VI. Conclusion and future research
Complete installation, e.g., Certification Authority (CA) and associated acts, is the basis of e-record development. Given the rapid progress of information technology, competition of medical industry, and national health insurance, the research of information security for e-record is a must. This paper applies the theory of group digital signature [4]
to Personal Digital Assistant (PDA) and smart card.
For users, our scheme is friendly and marketable. In the future, Medical Information Exchange Center (MIEC) should focus on secret sharing to prevent the loss resulting from the mistake or negligence of medical information holders.
References
1. Bemmel, J. H., Rotterdam and Rotterdam, M.A.
1999. Handbook of Medical Informatics.
2. Camenisch, J. and M. Stadler, 1997. ”Efficient group signature schemes for large
groups, ”Advances in Cryptology- CRTPTO’97, pp.410-424.
3. Cavallar, S., B. Dodson, A. Lenstra, P. Leyland, W. Lioen, P. L. Montgomery, B. Murphy, H. te Riele, P. Zimmermann, 1999. “Factorization of RSA-140 Using the Number Field Sieve,”
Advances in Cryptology-ASIACRYPT’99.
4. Chaum D. and E. van Heyst, 1991. “Group signatures,” Advances in Cryptology-EUROCRYPT’91 pp.257-265.
5. Chen, L. and T.P. Pedersen, ”New group signature schemes,” Advances in Cryptology- EUROCRYPT’94, pp.171-181, 1994.
6. Chen, J. J. R. and Y. Liu,2000. ”A traceable group signature scheme,” Mathematical Computer Modeling, pp.147-160.
7. ElGamal, T. 1985. “A public key cryptosystem and a signature scheme based on discrete logarithms,” IEEE Transactions on Information Theory, vol.31, No.4, pp.469-472.
8. Feige, U., A. Fiat, and A. Shamir, 1988.
“Zero-knowledge proofs of identity,” Journal of Cryptology, vol.1, pp.77-94.
9. Huber, G. P. 1984. “Issues in the Design of Group Decision Support System,” MIS Quarterly (8:3), pp.195-204.
10. Hwang, H. G. 2002. ”Research Issues In Healthcare Information System,” Journal of Information Management vol.9, Special Issues February.
11. Jian, W. S., Y. C. Li, D. D. Tang, and C. H. Hu, 1997(Dec) “Building a Prototype of Medical Information Exchange Center,” The Journal of China Association for Medical Informatics No6, pp.54-66.
12. Lorenzi, N. M. and R. T. Riley, 1995.
Organizational Aspects of Health Informatics.
13. Rivest, R., A. Shamir, and L. Adleman, 1978.
“A method for obtaining digital signatures and
4 public-key cryptosystems,” Communications of the ACM, Vol.21, No.2, pp.120-126.
14. Tseng, Y. M. and J. K. Jan, 1999. “Improved group signature scheme based on discrete logarithm problem,” Electronics Letters, Vol.35, No.1.
15. Wang, D. W., D. R. Liu and Y. C. Chen, 1999(Dec). “A Mechanism to Verify the Integrity of Computer- based Patient Records,”
The Journal of China Association for Medical Informatics, No.10, pp.71-84.
The Implementation of a Total Hospital Information System:
A Case Study in Selayang Hospital
Mohamad Azrin Zubir
a, Rosnah Hadis
a, MD, Isa Samat
baSelayang Hospital, 68100 Selangor, Malaysia
bInformation Technology Faculty, MARA University of Technology
Abstract
Selayang Hospital is the first hospital in Malaysia to implement a Total Hospital Information System (THIS). The aim of THIS was to provide a paperless and filmless hospital environment. The hospital has now been successfully operational for nearly three years with THIS as a major source of clinical and non-clinical information for healthcare providers.
This paper reviews in brief the challenges faced, implementation plan, system functionality, and the project outcomes, including results of a user satisfaction survey and early findings of an ongoing benefit study. Despite some growing pains, Selayang has ironed out many critical issues during the initial three years. The hospital recognizes it still has much to learn, improve, and share with others as it works to achieve its aim of a paperless and filmless solution.
Keywords
Hospital Information Systems, Systems Integration, Computer-based Patient Records, Implementation, Project Outcomes
1. Introduction
Selayang Hospital is the first hospital in Malaysia to implement a Total Hospital Information System (THIS) with the aim of providing a paperless and filmless environment. THIS started operation in August 1999; as of 5/6/2002, it had registered 150,000 patients and stored about 160 Gigabytes of electronic medical record in its clinical data repository. Located in suburban Kuala Lumpur, the 960-bed hospital provides specialized services to a 700,000-catchment population. The nearest hospital is located about 15km away.
Previously, the information systems used in the hospital were often developed and implemented independently at a departmental level.1 The lack of standardization combined with decentralized system development to create difficulties when there was a need to integrate the data from different ancillary systems. The result was redundant information, which was not cost effective. Addressing this issue, THIS was created to maximize the efficiency of information management in Malaysian government hospitals. The approach was to structure THIS so as to provide integration and interoperability.
As a pilot electronic hospital project established by the Malaysian Ministry of Health (MOH), Selayang Hospital is in essence a test-bed for the country.
Implementation of THIS was launched based on a business case focused on quantifiable benefits and intangible assets of improved information management in a hospital.
According to Shortliffe and Blois, “Implementing electronic records is inherently a systems -integration task; it is not possible to buy a medical-record system for a complex organization as an off-the-shelf product.
Joint development is crucial.”2 In Selayang, THIS is comprised of clinical, imaging, and administrative and financial information systems. These are fully integrated and interfaced through a Health Level 7 and DICOM standard protocol. The aim of any hospital information system is to support all affiliated hospital activities through a centralised database where information is stored, process, analyzed and shared across the hospital.3 Building such a complex IT environment in a healthcare organization was not an easy task. The daily clinical routine has undergone major process re- engineering and changing the mindset of healthcare providers towards IT processes.
Newly constructed, the hospital is wired with a local area network solution using asynchronous transfer mode (ATM) switching and fast Ethernet technology.
The network has a redundant fiber optic backbone that transfers packets of data using physical cabling and wireless infrastructure capabilities. Al l patient data and information can be retrieved at any of 1200 personal computers or 22 wireless notebooks in the hospital. Healthcare providers do the data entry themselves at the point of care.
2. Challenges
Building the first paperless and filmless hospital was a challenging project for the Malaysia MOH. First, there was a need to persuade healthcare providers to change their mindset towards using information technology in daily clinical processes. Second, MOH had no experience with complex information technology implementation in a hospital. Early in the project, a group of senior executives went abroad to identify a reference site but came back empty handed.
Third, MOH had to implement the framework without having proficient workers and planners in healthcare informatics field. By far the greatest
6 challenge, however, was the change management process. Most healthcare providers were used to paper-based records, and the computerized patient record had never been practiced in Malaysia.
3. Implementation Plan
The building of Selayang Hospital began in February 1996 and was completed February 1999. Information technology (IT) development was part and parcel of the physical building project. A contractor was appointed as a turnkey developer, but the MOH closely monitored the plan for and development of the building as well as IT implementation. A core team of senior executives, physicians, surgeon, nurses and allied health professionals oversaw the project direction and status. One of the team’s most important tasks was to embark on the Business Process Re -engineering (BPR) of all clinical and non-clinical process of the hospital. BPR concepts are widely recognized in both academic and business literature4-7 and are usually associated with the management of the transition of an organization from traditional processes to the use of information technology. The outcome of various BPR meetings produced a consensus basis for improved workflow and changes in standard hospital operating procedures and policies.
To ensure that the project was clinically driven rather than vendor pushed, the core team involved many healthcare providers from its very early phases. This early involvement created champions in every department who later helped to influence other personnel to accept the new working lifestyle. These champions also helped the core team design the user interfaces and the fields for capturing data needed by their respective departments.
4. System Functionality
In THIS, the hospital acquired a best-of-breed off-the-shelf product with some customization.
During the initial project phase, there were no systems that were able to provide an integrated solution of computer-based patient record (CPR).8 Hence, in Selayang three major systems— Clinical Information System, Picture Archiving Communication System (PACS), and Administration and Finance System— were integrated in a design developed to fulfil the workflow requirements of the hospital and support patient-centered care.
These three major systems and other systems, as shown in Table 1, were interfaced according to HL7 standard format using an open interface engine. All the clinical data are stored in a central data repository.
With a separate data repository, the administration and finance system is linked with the clinical system only through billing system for patient accounting.
The PACS stores all radiology images in an imaging repository using jukeboxes.
Module Provider Function Person
Management
Cerner Data entry for registration.
Inpatient/outpatient admission transfers and discharge. Bed management system.
Clinical Documentati on
Cerner Structured data entry for history, physical examination,
treatment plan, nursing tasks, case summary and referral.
Order Management
Cerner Computerized order entry systems for care providers. Includes all orderables available in the hospital.
Scheduling Cerner Appointment entry, process and display for specialist clinic, procedure room, rehabilitation, daycare center and operating room schedule. Scheduling management system for clinics, operating room, and daycare center.
Laboratory Information System
Cerner Processing request from physician, sample management, analytical processing via interfacing with lab equipment, duplicate checking and electronic reporting. Multiple submodules, e.g., blood bank, pathology.
Radiology Information System
Cerner Processing request from physician, request management, duplicate checking and electronic report generation of images.
Picture Archiving and
Communicat ion System
Siemens Image processing, film archiving, tracking, distribution of digitized images throughout the hospital via the network and image
7 display.
Critical Care Monitoring System
Spacelab Data capture from patient monitoring devices, fluid management in critical care unit.
Calculation of drug formulary.
Pharmacy Information System
Cerner Only outpatient drug prescription process available. Full functionality to come in mid 2003
Table 1: System Functionality
5. User Satisfaction
A survey of emergency room physicians at Selayang in early January 2002 revealed widespread of approval of THIS. According to the survey,
• 71 percent of emergency room physicians found Cerner’s Powerchart, the user interface for the emergency medical record, excellent or very good in terms of ease of use.
• 71 percent rated the process of ordering investigations through THIS excellent or very good.
• 64 percent rated the process of viewing test results through THIS excellent or very good.
• 64 percent rated viewing patient information through THIS excellent or very good.
• 71 percent rated viewing information through the form browser excellent or very good.
• 85 percent rated system data integrity very good or average.
• 78 percent rated system data security very good or average.
• 93 percent said they preferred working in a paperless environment vs. a traditional paper-based hospital.
Physician respondents gave other reasons for preferring THIS: having patient records at fingertips;
easy retrieval of patient records; faster results; ease of scheduling; convenience; efficiency; security; no paper forms; no difficulty reading poor handwriting;
quicker access to old notes.
In the area of improving performance, the physicians suggested:
• Improved record security
• Less hanging time
• Improved help desk service
• Improved system speed.
6. Benefit Study
A comprehensive study quantifying the benefits generated by THIS is underway, with the first published results expected late in 2002. To date, there have been several focused studies, including one comparing Selayang Hospital and a similarly sized, non-THIS hospital in Malaysia called Tengku Ampuan Rahimah Hospital. Table 3 presents some initial results.
Benefit Description 36.7 % time
reduction during admission
Admission Documentation.
Selayang’s process takes 1 minute and 5 seconds less per patient than Ampuan Rahimah’s.
8.3% time reduction during the patient’s stay
Clinical Documentation. Selayang’s daily activities take 40 seconds less per patient than Ampuan Rahimah’s.
71.1% time reduction at the patient’s discharge
Discharge Documentation.
Selayang’s process takes 4 minutes and 20 seconds compared to 15 minutes at Ampuan Rahimah.
73% reduction in time reporting following a portable X-ray
Selayang takes 14 minutes to take an X-ray with a portable devi ce and post the report in the CPR compared to 52 minutes at Ampuan Rahimah to take the images and send the paper report to the relevant ward.
80% reduction in time to view the patient’s medical record
Selayang takes the same amount of time to trace an entire paper-based case note as Ampuan Rahimah takes to retrieve the last paper data from the patient’s last visit. At Selayang, computerized patient record users access within 1 minute all the data that is stored about a patient online. At Ampuan Rahimah, a complete historical record for patient is not routinely provided.
11.32%
reduction in staff for patient registration and scheduling functions
Selayang requires 94 employees to perform patient registration and scheduling processes, compared to 106 employees at Ampuan Rahimah, who continue to rely on traditional manual processes.
51.6%
efficiency gain in scheduling patients
Selayang’s 94 registration and scheduling employees can schedule 238 appointments in the time it takes Ampuan Rahimah’s 106 registration employees to schedule 157 appointments.
Table 2 : Benefits of THIS
8
7. Lessons Learned
The list of lesson learned in Selayang Hospital is long indeed. Table 2 summarizes some important lessons from the first three years of implementation.
Area Lesson Learned
Planning • Focus on seamless integration and patient-centered flow
• Involve end users early to increase technology diffusion
• Foster partnership between all vendors and healthcare providers
Business Process Re-engineering
• Critical to achieve in the early phase
• Takes most of the project time
Operation • Information system needs fast response time due to massive clinical data
• Support must be efficient and understand medical urgency
• Never underestimate hardware and network requirements
• Knowledge in healthcare informatics is important
• Training of personnel in information and communication technology is critical
Information System
• Minimize interfacing where possible but maximize integration
• Standardization for capturing data is important
• Decision and consensus on information flow
Server Room • Redundancy of all servers
• Internal mirroring
• Off-site replication of server room (data recovery plan)
• Redundant power supply unit Contractual
Issues
• Analyze and minimize all risk factors and possibilities prior to signing
Table 3: Lessons Learned
As the survey results suggest, the implementation of THIS, particularly the CPR, was a complex undertaking. Selayang has learned valuable lessons from its experiences, including lessons from system mistakes, errors, user complaints, and failed implementations in certain areas. Development over the next three years will focus primarily on improving the completeness and quality of CPR.
8. Other Assets
In addition to these quantifiable benefits, Selayang Hospital has identified other assets which it hopes will result from implementing THIS. These include the following:
• Creation of a knowledge based culture through user acceptance of information sharing
• Provision of data warehousing opportunity for data mining, ease of statistic reporting and research
• Improved care delivery efficiency
• Improved patient safety
• Creation and streamlining of a single lifetime health record for the individual patient.
9. Conclusion
Selayang Hospital provides a compelling case study of the implementation of a paperless and filmless hospital information system. Studies to date have documented user satisfaction and quantified system benefits. Future plans promise to improve the quality and completeness of computer based patient record and provide detailed cost benefit analyses. It is hoped that these studies will also aid in defining assets that have proved difficult to document in the past, including patient safety and areas affecting quality.
References
1. C Safran, LE Perrault, Management of information in integrated delivery networks, in: EH Shortliffe, LH Perreaulll, G Wiederhold, LM Fagan (Eds.) Medical Informatics: Computer Applications in Healthcare and Biomedicine, Springer, New York, 2001, pp. 359-396.
2. EH Shortliffe, MS Blois. The computer meets medicine and biology: emergence of a discipline, in:
EH Shortliffe, LH Perreaulll, G Wiederhold, LM Fagan (Eds.) Medical Informatics: Computer Applications in Healthcare and Biomedicine, Springer, New Yo rk, 2001, pp. 3-40.
3. H Lodder, AR Bakker. Hospital information systems: technical choices, in: JH van Bemmel, MA Musen (Eds.) Handbook of Medical Informatics, Springer, New York, 2000, pp. 343-356.
4. V Grover, WJ Kettinger (Eds.), BPR:
Reengineering Concepts, Methods and Technologies, Idea Publishing Group, London, 1995.
5. H Osterle, Business in Information Age: Heading for New Processes, Springer, Berlin, 1995.
9 6. C Coulson-Thomas (Ed.), Business Process Re-engineering: Myth and Reality, Kogan Page, London, 1994.
7. N Venkatraman, IT-enabled business transformation: from automation to business scope redefinition, Sloan Management Review, Winter 1994, pp. 73-87.
8. RS Dick, EB Steen, DE Detmer (Eds.), The computer-based patient record: an essential technology for health care, revised edition, National Academy Press, Washington, DC, 1997.
___________________________________________
Correspondence:
Dr Mohamad Azrin Zubir Hospital Selayang,
Lebuhraya Kepong Selayang, 68100 Batu Caves
Selangor
10
Electronic Health Record in Cardiology:
Pilot Application in the Czech Republic
Jana Zvarova, Petr Hanzlicek, Josef Spidlen
European Center for Medical Informatics, Statistics and Epidemiology – Cardio, Institute of Computer Science, Academy of Sciences of the Czech Republic
Pod Vodarenskou vezi 2, 182 07 Prague, The Czech Republic http://www.euromise.org/
Abstract
The EuroMISE Centre -Cardio is focusing on new approaches to electronic health record, considering also the Czech healthcare environment.
The development of new approaches is based on experience gathered in I4C and TripleC projects of the European Union. In the paper we discuss basic requirements on electronic health record and special needs of the Czech healthcare system. Finally, pilot application named MUDR implementing proposed concepts is described.
Keywords:
Electronic Health Record, Health Information System, Biomedical Informatics
1.
Introduction
The European Center for Medical Informatics, Statistics and Epidemiology – Cardio (the EuroMISE Center – Cardio) is the joint workplace of five institutions. The Center associates more then 70 researchers from the Institute of Computer Science, Academy of Sciences of the Czech Republic, Charles University, University of Economics, University Hospital in Prague and Municipal Hospital in Caslav.
The EuroMISE Center is focused on new approaches of designing and using of electronic health record (EHR) including telemedicine applications, intelligent systems for data mining and decision support [1], electronic medical guidelines [2] and PhD education in biomedical informatics [3].
The EuroMISE Center is running two ambulances of preventive cardiology. Further we introduced more details on MUDR (MUltimedia Distributed Record) development in the EuroMISE Center and its pilot applications in ambulances of preventive cardiology.
2.
EHR development in the EuroMISE Center
The research on EHR for data acquisition, data
storage and data mining has been carried in the EuroMISE Center. Different databases, storing clinical, genetic and epidemiology data, were created.
It was found that data collected in cardiology had different quality and accuracy. Minimal data model for cardiology was proposed to assure basic cardiology data collection with high quality and accuracy over a long period of time focused on further improvement of cardiovascular diseases management [4].
Cardiology databases can serve for comparison of medical practice with medical guidelines and discover some features of diseases that can help to their management. Medical knowledge in different medical guidelines can be repeatedly evaluated using large amount of data collected by EHRs and stored in a cardiology database. The large databases in cardiology, based on appropriate data model, can further serve as a source of information for decision support systems and can be explored by data mining methods to reveal hidden associations and relationships. Moreover, EHR serving as a tool for data collection can be used for automated generation of alerts, reminders and suggestions when standards of care (e.g. based on medical guidelines) are not achieved.
Development of the EHR architecture at the EuroMISE Center was inspired by several European projects [5], mostly by the I4C and TripleC projects. In the I4C project the "Multimedia Electronic Patient Record" named ORCA (Open Record for CAre) was developed [6]. The software was based on 2-layer architecture (database and user interface level), integrated structured data entry combined with the possibility to include multimedia objects as part of patient history. The primary user interface of ORCA was the structured data entry, allowing adding free-text comments to entered data.
The data set of collected data was defined by hierarchically structured knowledge base, described in many European languages. Entered data were therefore easy to translate to different languages even if the other language was used during entry.
The knowledge base editor allowed
the administrators to modify the set of collected data.
The EuroMISE Center participated in the TripleC project. Among the goals of this project was the adaptation of the ORCA software to Czech environment and testing of usability of structured EHR in Czech hospitals [7]. The ORCA system introduced many new and useful ideas and principles, however many features needed for Czech healthcare environment were missing. The system of structured data entry used in ORCA showed to be more time consuming than the free-text entry and therefore not preferred by some Czech physicians.
3.
Required characteristics of EHR
EuroMISE centre experience in different medical informatics and statistics applications, ideas inspired by results of I4C and TripleC projects and long time cooperation with physicians resulted in the following list of requirements on EHR.
• Structured way of data storage combined with free text;
• Tools to ease structuring of entered information (transfer from free-text to structured data);
• Tools for evaluation and visualisation of stored data;
• Dynamically extensible and modifiable set of features without change of database structure (knowledge base);
• System for access control to patient data and knowledge base;
• Module checking data for correctness and conformance;
• Multilinguality;
• Pedigree information;
• Grouping of patient data according to cases;
• Logging system maintaining information on every change of patient data, knowledge base, access rights modification etc.;
• Minimal dependence on database and operating system used for data storage;
• Wide scale usability (from single workstation used by general practitioner to distributed environment in a large hospital);
• Mobile terminals interface (PDAs, mobile phones);
• Motivation systems for physicians (routine administrative work automation, reports for insurance companies generation, etc.);
• Multimedia information as the part of EHR.
4.
Architecture of the MUDR
The preparation phase of the MUDR development consisted of international standards analysis, modelling of the architecture of the whole system, process and functional analysis, data model preparation, etc. As a part of the preparation phase of the software development, international standards were analysed. Several organizations producing
European and international standards exist nowadays.
At the European level, the technical committee 251 of the European Committee for Standardization (CEN/TC251) produced over 40 preliminary European standards in the field of health informatics.
The main inspiration for our work is the document ENV13606 "Electronic healthcare record communication". The prime purpose of this multipart pre-standard is Electronic Healthcare Record communication; communication being defined as the act of imparting information. The first part describes a conceptual model of structure and content suitable for communicating EHR [8]. Implementation of the described model in our system was considered.
Following the requirements stated before, the modular structure of the system was defined.
The main architecture of MUDR record is developed using 3-layer architecture - database layer, the application layer and the user interface layer. This approach separates the physical data storage, the application intelligence and the user interface and minimizes the requirements to the client side software. Two versions of database and application layer are developed - the universal one, using relational database structure and the special one using object-relational database structure of Oracle 9i.
The comparison of efficiency of the classical relational architecture and the new object-relational architecture will be performed.
As stated in the list of required characteristics of EHR, to be able to use universal technique for processing and visualisation of stored data we need unified way of structured data storage. The set of collectable features varies in different departments, organizations and also during time. Therefore we need dynamically extensible and modifiable structure of items allowing reorganization without change of database structure. The main idea behind our database implementation is the separation of data values and data description.
5.
Database representation
The set of features and their relations is described by graph structure that consists of vertices representing features edges representing relations among features.
The basic structure underneath is the oriented graph (tree) representing the hierarchy of medical knowledge stored in the health record by edge type
"subjection". Other types of edges are used to express relations such as equivalence of some terms in different parts of knowledge tree. We call this tree structure "knowledge base". Each vertex describes the represented feature by its identification, internal name, type and identifications of user who created the vertex and user who eventually deleted the vertex.
Similarly are described the edges between vertices.
This enables the system to log any change in the structure. If the structure is changed for any reason, the "deleted" vertices and edges are only marked as deleted and identification of the user who
12 deletes them is stored in them. The system enables the administrator to define access rights to any part of the tree to individual users or group of them similarly as access rights to directory structure in UNIX system.
The graph structure can also be used to express relations between patients in the database, e.g.
pedigree information. The graph is also used for description of usable scales and classifications. An example of hierarchically structured classification system is International Classification of Diseases (ICD10) or Anatomic Therapeutic Chemical classification (ATC). Both mentioned classification systems are already included in the implemented system. The internal name of the vertex is then used for unique identification of the vertex in the hierarchy by specifying the path from the root of the tree (e.g.
SCALES.MKN10.V.F60-F69.F63.2). Names of features in different languages, displayed in user applications, are stored separately.
The collected data are stored using similar structure as knowledge base. Each vertex in the tree describes one instance of the feature from the knowledge base by the identification of the feature (internal name of the feature), its value (with the possibility to specify the range of values), date and time of examination, date/time range of the validity of determined data, certainty of the determined data and identification of user who entered, confirmed (doesn't have to be the same as the person who entered the data) and eventually deleted the instance.
Fig. p: Graph structure of the knowledge base and patient data
The values are physically stored in separate tables according to their types and are related to their descriptions in the tree. This simple data structure allows storing almost any type of structured information in the same way, making possible to apply generic algorithms to extract any type of information (e.g. time progress of laboratory tests, blood pressure, etc.).
The application layer implements the XML based communication protocol with the MUDR system.
The communication syntax between client and application layer is defined by XML Schema - recommended by W3C in May 2001. The XML documents are transported over HTTPS protocol between client application and HTTP server using CGI scripts. The CGI scripts communicate with the main application, which translates the XML
commands or queries to SQL queries and responses from SQL server to XML documents. The application logic offers basic commands for knowledge base editing, storing and recalling the patient data, modifying access rights etc. Important part of application layer is the guidelines module, which can serve as a decision support tool during the patient examination or as a compliance analysis tool.
Currently developed version of application layer runs as MS Windows NT service, guidelines modules are prepared as DLL libraries, communicating with the application layer. In the future, more platform independent solution for application layer (Corba, Enterprise JavaBeans, etc.), SOAP based communication and universal guideline module is considered.
6.
User interface
The user interface of described system can be developed in many ways using various technologies [9]. As stated before, the free text patient record, preferred by many physicians because of its speed and flexibility is very difficult for automated analysis.
To be able to statistically process the entered data, the structured record is essential. Therefore the user interface should help users of the system to create the structured data while keeping the speed of free text entry. One of the possibilities studied was the regular analysis of the free text reports and automatic generation of records to the patient database. Another possibility is based on idea of continuous offer of features from the knowledge base related to written text during entry of free text record.
User entering the report has the possibility to pick the feature from the offered list and to enter its value, range of validity and certainty. The system can then generate the sentence containing entered data and insert it into free text record. Such sentence is marked as related to specific item in the database of collected data. Manual mark-up definition of parts of entered text to items in database is also possible. The entered and marked free text report is then stored in database as XML document, giving physician the possibility to use his language to describe the examination of patient and defining links to items in the database for further processing of the data.
However, still the big problems occur with handling of EHR when the most of information is stored in the textual form. In this context we focus on the application of regular semantic analysis of the text medical reports that can lead to structured form of the information. A new system SFT (Semi Free Text) has been proposed [10]. It can store patient data in both forms, structured form and free–text at once. The SFT system implementations are based on the implementation of the knowledge base that defines the hierarchical structure of information on patients. The data are stored in the database in a form that allows both views, free–text and structured form view. The part of
the SFT system is also the module for the automatic analysis of the free text using regular grammars. It is a big advantage that this method is not bounded with a particular language. Moreover, it is not so much time consuming and there is no need of other large data sources, e.g. vocabularies. Therefore the semantic marking of medical reports is an algorithm that provides semantic analysis according to given rules and marks (e.g. using XML tags) the founded fragments of the medical report with a reference to the given knowledge base.
7.
Implementation
The pilot implementation of the proposed electronic health record structure is developed in the field of cardiology. The knowledge base is described by so called minimal data model for cardiology - the hierarchically organized list of symptoms from the field of cardiology. The minimal data model was developed by consensus of physicians and computer scientists cooperating in the EuroMISE Center in the field of electronic health documentation. A simple application for collection of data described in the minimal data model was prepared until the final form of MUDR is developed [11]. The data collected by this way will then be imported to the final form of the MUDR, compatible with the following Czech healthcare standards.
Czech unique identification number of the structure YYMMDD/XXXX, which uniquely identifies each citizen. The number is used in all health institutions, pharmacies, health insurance companies and public administration;
List of healthcare procedures with approximately 8000 health procedures accomplished not only with point values used for calculation of payments to health insurance companies, but also by time limits for running of services, that are used in special calculations (e.g. available at the server of Czech Ministry of Health - http://www.mzcr.cz).
International Classification of Diseases (ICD10) in Czech (MKN10), expressing the diagnosis as the four-digit code (see e.g. http://www.uzis.cz/).
Medication for a Czech healthcare provider allows prescribing drug from the list of produced drugs paid fully or partially by General Health Insurance Company of the Czech Republic (actualized quarterly). Therefore it is necessary to make a selection of drugs from MUDR directly.
Laboratory data are transferred according to Data standard for transferring of data on patients among healthcare institutions. Further Czech health providers are using National classification of laboratory items to communicate laboratory results.
Acknowledgment
The work was partially supported by the Ministry of Education of the Czech Republic by the grant no.
LN00B107
References
[1] Rauch, J: Mining for Statistical Association Rules. In The Fifth Pacific/Asia Conference on Knowledge Discovery and Data Mining Industrial track and Workshop Proceeding Red.
Joseph Fong and Michael Ng Hong Kong 2001, 149-158
[2] Zvárová, J, Peleska, J, Hanzlícek, P, Zvára, K:
Enhanced Care of Hypertensive Patients Using Internet. Technology and Health Care 9, 6, 2001, 487-488
[3] Zvárová, J, Svacina, S: New Czech postgraduate doctoral program in biomedical informatics.
Proceeding of MIE 2002 (in print)
[4] Mannsmann, U, Taylor, W, Porter, P, Bernarding, J, Jager, HR, Lasjaunias, P, TerBrugge, K, Meisel, J. Concepts and Data Model for a Co-Operative Neurovascular Database. Acta Neurochirurgica 143, 2001, 783-791
[5] Iakovidis, I.: Towards personal health record:
current situation, obstacles and trends in
implementation of electronic healthcare record in Europe. Int J Med Inform 52, Nos.1 -3, 1998, 105 –115
[6] Pierik, PH, van Ginneken, AM, Timmers, T, Stam, H and Weber FR, Restructuring routinely collected patient data: ORCA applied to
andrology. Methods Inf Med 36, 1997, 184-190 [7] Zvárová, J, Hanzlícek, P and Pribik, V,
Application of ORCA multimedia EPR in Czech hospitals. In: EuroREC 99, ProRec, Sadiel, Sevilla 1999, 160-165
[8] ENV13606 Parts 1-4, CEN/TC251 [9] van Ginneken AM, Verkoien, M.J.
A multi-disciplinary approach to a user interface for structured data entry. MEDINFO 2001, 693-697
[10]Semecky, J and Zvarova, J: On regular analysis of medical reports. In R. Baud and P. Ruch (eds) Natural language in biomedical applications, Nicosim Cyporsu 2002, 13-16
[11]Mares, R, Tomeckova, M, Peleska, J, Hanzlicek, P, Zvarova, J: Uzivatelska rozhrani pacientskych databazovych systemu - ukazka aplikace pro sber dat v ramci minimalniho datoveho modelu kardiologickeho pacienta, Supplementum Cor et Vasa 44, 4, 2002, 76
Address for correspondence:
Jana Zvarova, EuroMISE Center
Pod Vodarenskou vezi 2, 182 07 Prague 8 The Czech Republic
email: [email protected]
14
Reliability of Dermatologic Teleconsultations with a Web-based Technology
Kemal Hakan Gülkesen
a*, Murat Ö ztas
b, Emel Çalikoglu
c, Kiymet Baz
d, Ahu Birol
e, Meltem Ö nder
b, Tamer Çalikoglu
f, Mehmet T. Kitapçi
gaBiostatistics Department, Medical School, Akdeniz University, 07059 Antalya, Turkey
bDermatology Department, Gazi University Medical School, Ankara, Turkey
cDermatology Department, Fatih University Medical School, Ankara, Turkey
dDermatology Department, Mersin University Medical School, Mersin, Turkey
eDermatology Department, Kirikkale University Medical School, Kirikkale, Turkey
fRadiation Oncology Department, Oncology Hospital, Ankara, Turkey
gNuclear Medicine Department, Gazi University Medical School, Ankara, Turkey
* The corresponding author [email protected]
Abstract
Teledermatology applications are based on two different modes of technology, store and forward, and real time. We designed a study to test the reliability of teledermatologic diagnosis by a web-based store and forward system.
Clinical information and photograph(s) for each of previously diagnosed 125 dermatology patients were placed on a web server. Three teledermatologists had tried to make single most likely diagnosis by the help of a web interface. They also had an option of not to make any diagnosis.
Diagnostic accuracy rates of three teledermatologic consultants were 0.62, 0.69, and 0.80 and these rates statistically differ (p<0.05). Agreement rates between teledermatologists were 0.55, 0.70, and 0.57. 76.8 % of the patients could be correctly diagnosed by two or three of the consultants.
These results suggest that a web-based store and forward system may be reliable for teledermatologic consultation, and only one well-oriented teledermatologist has higher accuracy rate than combined view.
1. Introduction
Teledermatology applications are based on two different modes of technology, store and forward, and
real time. In store and forward technology, photographs of patient are sent and can be reviewed by consultant at any time. Real time consultations use videoconferencing technology, and permit the interaction of teledermatologist and patient. Use of store and forward technology seems to lead more hospital referrals [Loane, 2000]. On the other hand, the main disadvantage of real time technology is its cost. There is a continuing debate on the selection of the system [Whited, 2001].
Conventional store and forward systems need a digital camera, at least two computers which contain the required software, and communication between these computers. We designed a study to test the accuracy and reliability of teledermatologic diagnosis based on a web-based store and forward system. A web-based system can be used in any computer, which is connected to Internet, so the users do not need to use certain computers in certain place, or install software. In our study design, to test the importance of clinical information, the teledermatologist had to make two separate diagnoses, without clinical information, and with clinical information.
Table I: The diagnostic accuracy rates of three teledermatologists (A, B, and C) with and without clinical information. Accuracy T: Accuracy in total cases. Accuracy D: Accuracy when teledermatologist makes a diagnosis. CI: Clinical information.
No diagnosis False diagnosis Correct diagnosis Accuracy T Accuracy D
CI (-) (+) (-) (+) (-) (+) (-) (+) (-) (+)
A 18 (14.4 %) 9 (7.2 %) 38 (30.4 %) 30 (24.0 %) 69 (55.2 %) 86 (68.8 %) 0.55 0.69 0.65 0.74 B 18 (14.4 %) 14 (11.2 %) 37 (29.6 %) 34 (27.2 %) 70 (56.0 %) 77 (61.6 %) 0.56 0.62 0.65 0.69 C 21 (16.8 %) 6 (4.8 %) 28 (22.4 %) 19 (15.2 %) 76 (60.8 %) 100 (80 %) 0.61 0.80 0.73 0.84
2. Material and Method
125 dermatology patients were randomly selected from outpatients of Gazi University Dermatology Department. All the patients had a definite diagnosis based on clinical examination and/or biopsy. The lesions of all the patients were photographed using a digital camera (800x600 pixel, JPEG compression).
A web based software was developed by Pleksus Bilisim (Ankara, Turkey) using Apache 1.3.20, MySQL 3.22.32, PHP 4.0.3pl1. The software is composed of a database containing clinical information and photographs of the patients, responses of each teledermatologist for each patient, and web interfaces for the teledermatologists and the moderator. After placing the patient data by the moderator, the application was opened to use of teledermatologists on web server (http://dermatoloji.egitimi.org). Each tele-dermatologist was given a login and password to see patient data and enter her diagnoses. They had used 17-inch monitors and 1024x768 screen resolution. One of them was in Geneva, Swiss (EÇ), the others were in Mersin, Turkey (KB), and Kirikkale, Turkey (AB). All the teledermatologists were dermatology specialists who had been practising for two or three years.
At the first step, the teledermatologists were not given clinical information. They tried to make single most likely diagnosis based only the photographs.
They also had an option of not to make any diagnosis.
At the second step they could see the clinical information, and were able to change their diagnosis.
SPSS software was used for the statistical analysis.
Chi-square test was used to test the interobserver and Mc Nemar test was used to test intraobserver differences. Accuracy rates were calculated as the percent of true diagnosis in total cases. Interobserver agreement rates were calculated as the percent of agreement between two consultants irrespective of the diagnosis is true or false.
3. Results
Diagnostic accuracy rates of three teledermatologic consultants are presented in table I. Interobserver and intraobserver statistical comparisons of these accuracy rates are presented in table II.
Table II: Intraobserver difference and interobserver difference of accuracy rate. CI: Clinical information.
Interobserver difference
CI (-) (+)
Intraobserver difference AvB >0.05 >0.05
A
0.003AvC >0.05 0.042 B >0.05 BvC >0.05 0.001 C 0.001
Agreement rates between the teledermatologist are presented in table III. In table IV, the numbers of the cases according to number of the accurate diagnosis by the teledermatologists are presented. In other words, the cases were classified according to number
16 of correct diagnoses.
Table III: Agreement rates with and without clinical information. CI: Clinical information.
Agreement rate
Without CI With CI
A vs. B 0.46 0.55
A vs. C 0.47 0.70
B vs. C 0.44 0.57
Table IV: Distribution of correct diagnoses according to patients.
Correct diagnosis Without CI With CI 0 27 (21.6 %) 14 (11.2 %) 1 21 (16.8 %) 15 (12.0 %) 2 37 (29.6 %) 40 (32.0 %) 3 40 (32.0 %) 56 (44.8 %) Total 125 (100.0%) 125 (100.0%)
4. Discussion
Interobserver agreement rates in the literature are ranged from 0.54 to 0.96 using store and forward technology [Whited, 2001], and results in present study are within these limits (0.55, 0.57, 0.70).
However, agreement rates without clinical information are below the usual limit (0.44, 0.46, 0.47). As expected, clinical information is definitely required in teledermatology. Without clinical information, the diagnostic accuracy rates of teledermatologists are from 0.55 to 0.61, and there is no statistical difference between them. With addition of the clinical information, accuracy rates rise to 0.62-0.80, and statistical difference appears between the teledermatologists. In our opinion, the cause of this difference may be orientation problem with the teledermatology procedure. The accuracy is determined according to definite diagnosis in our study. In the medical literature, there is only one study comparing teledermatology diagnosis to definite diagnosis [Whited, 1999]. According to this study, accuracy rates range from 0.53 to 0.63 when accuracy was based on the single most likely
diagnosis. According to this result, web-based store and forward technology has an acceptable accuracy.
Only 76.8 % of the patients could be diagnosed by two or three of the consultants. This rate is lower than the rate of the teledermatologist who has an accuracy rate of 0.80. This comparison shows that instead of multiple consultants, a well-oriented consultant is better.
Our study indicates that web based store and forward technology is comparable to conventional store and forward methods, and may be a significant tool for Teledermatology applications. Our results also suggest that accuracy rates between teledermatologists may significantly differ. Only one well-oriented teledermatologist has higher accuracy rate than combined view.
References
1. Loane MA, Bloomer SE, Corbett R, et al. A randomised controlled trial to assess the clinical effectiveness of both realtime and store-and-forward teledermatology compared to conventional care. J Telemed Telecare 2000; 6 Suppl. 1: 1-3
2. Whited JD, Hall RP, Simel DL, et al.
Reliability and accuracy of dermatologists’
clinic-based and digital image consultations.
J Am Acad Dermatol 1999; 41: 693-702 3. Whited JD. Teledermatology, current status
and future directions. Am J Clin Dermatol 2001; 2: 59-64
18
20
A Pattern Recognition Incorporating Temporal and Spectral Features for Cardiac Arrhythmia Detection
Lung-Hsun Chang, Yi -Chen Lin, and Szi-Wen Chen
Department of Electronic Engineering, Chang Gung University, Kwei-Shan, Tao-Yuan 333, Taiwan [email protected]
Abstract
In this paper, a pattern recognition based algorithm developed for the discrimination of ventricular fibrillation (VF) and ventricular tachycardia (VT) is presented. The method jointly utilized temporal and spectral features, in conjunction with a linear discriminant function (LDF), for the task of VF/VT classification. Tests conducted using 70 ECG episodes collected from the MIT-BIH database were performed. The numerical results showed that the novel method achieved correct detection rates of 96.00% and 95.56% for VF and VT, respectively, permitting 95.71% overall detection accuracy.
1. Introduction
A number of previous literature/reports have indicated that cardiac arrhythmias account for over 50% of all deaths due to heart disease.
Prevention of such cardiac deaths requires rapid and accurate identification of lethal arrhythmias , such as ventricular fibrillation (VF) and ventricular tachycardia (VT), either from the surface or intracardiac electrocardiogram (ECG) tracings . Since VF and VT are required to be effectively treated by delivering different electrical energy shock at early stage, development of built-in algorithms f o r “smart”
defibrillators that are capable of differentiating VF from VT as well as making quick and accurate shock decisions is thus essential to reducing mortality from cardiac deaths. There were a number of detection algorithms for cardiac arrhythmias having been proposed
previously [1], [2], [3].
In this paper, a novel method based on time and frequency analysis is introduced. Two features, dubbed threshold crossing interval (TCI) and frequency spread, were defined in our study.
In general, after obtaining both features respectively by applying time and frequency analyses to the original ECG recordings, an optimal linear mapping procedure was then applied. Such a linear mapping is adopted to reduce a 2-D feature vector to a 1-D scalar so that the detection of VF and VT can be achieved simply by comparing the sole feature value with a pre-determined threshold that well separates VF from VT. Descriptions of the method and performance evaluation are given in the subsequent sections.
2. Method
A block diagram of the overall process flow is depicted in Fig. 1. Initially, an ECG signal is preprocessed by a bandpass filter with the passband of [2,20] Hz to remove the baseline drift, the motion artifacts and the 60Hz power line interference.
A. Feature Extraction
A time-domain feature was defined as follows.
First, a 5 s ECG signal was converted into a binary sequence by comparing the ECG sample value with an adaptive threshold. If a sample value was greater than or equal to the threshold, then a digit “1” was assigned to represent that sample; otherwise, a digit “0” resulted. Note that the threshold was selected as 20% of the peak
22
Figure 1.Block diagram of the process flow
value of a 1 s ECG segment and was updated every 1 s, thus producing an adaptive threshold. As a result, a number of pulses were generated. Such a thresholding scheme can reduce low-level noise and does not affect the heart signal. Moreover, adaptive thresholding leads to immunity from significant changes in the signal amplitude. Next, the number of pulses, denoted as N, was counted. The mean pulse interval, referred to as the threshold crossing intervals (TCI’s) corresponding to a binary sequence resulting from the original 5 s ECG signal can be estimated by TCI=5000/N (ms). TCI was then considered as the temporal feature in our method.
As for the spectral feature, it was defined in the following manners. The fast Fourier transform (FFT) was first computed for each original 5 s ECG signal.
Fig. 2 gives a typical plot of the magnitude frequency response obtained from a VF ECG recording. We here took the 50% of the maximum magnitude as the
“clip” and denoted the largest and the smallest frequencies whose corresponding spectral magnitudes were greater than the clip value as fh and fl, respectively. A feature referred to as frequency spread, denoted by S, was then defined as
S= fh - fl,
where S has the unit in Hz.
Figure 2. Spectrum of a VF ECG signal
B. Pattern Classification
To perform the task of classification, two features were jointly applied. A linear mapping derived based on a likelihood ratio test (LRT) was introduced for this purpose and formulated as:
=
S A TCI
y
T ,where A is a 2×1 linear mapping matrix used to transform from a 2-D feature vector to a smaller 1-D scalar y that retains the information relevant for classification [4]. In our study, the
matrix A was derived from an existing training database. As a result, the decision rule can be expressed as
If y≧T, => VF;
If y﹤T, => VT,
where T represented a threshold that can well differentiate VF from VT groups.
3. Results and Discussion
A subset of the MIT-BIH database consisting of 70 recordings (VF: episode numbers 1-25; VT: episode numbers 26-70) was adopted in our analysis. Each recording was 10 s in length (the sampling frequency fs = 250Hz). Consequently, the optimal results achieved by our algorithm were 96.00% (24 out of 25) for VF detection and 95.56% (43 out of 45) for VT detection, producing 95.71% overall detection accuracy, as highlighted in Table 1. A scattering plot of the classification results was also given in Fig. 3.
Table.1 Performance with different threshold values.
The optimal result is highlighted.
Finally, several points are addressed as follows.
First, when either time or frequency feature was solely applied, the performances resulting from both cases were both lower than that obtained from a joint use of both features, implying that the information
utilized for classification in time and frequency domain should be independent each other. Secondly, it should be noted that one limitation of our study is that the algorithm performance was tested on the same database for training process. Therefore, a larger database would be essentially required for prospectively evaluating the pattern classification performance for our method. Moreover, since the novel algorithm only required a short ECG recording (say, 5 s, for example), thus it can meet the requirements, such as rapid and accurate shock decision, for implementation in a practical automated external defibrillator
(AED)
Figure 3. Plot of classification (VF:x; VT:o)
4. References
[1] N. V. Thakor, Y. -s. Zhu, and K.-Y. Pan,
“Ventricular tachycardia and fibrillation detection by a sequential hypothesis testing algotithm,” IEEE Tran. Biomed. Eng., vol. 37, pp. 837-843, 1990.
[2] S.-W. Chen, “A two-stage Discrimi- nation of Cardiac Arrhythmias Using a Total Least Squares-Based Prony Modeling Algorithm,”
IEEE Tran. Biomed. Eng., vol. 47, pp.
1317-1327,Oct. 2000
[3] M. E. Cain, H. D. Ambos, J. Markham, B. D.
Lindsay, and Arthur, “Diagnostic implications of spectral and temporal analysis of the entire cardiac cycle in patients with ventricular tachycardia,” Circ., vol. 83, no.5, pp. 1637-1648, May 1991.
[4] K. Fukunaga, Introduction to Statistical Pattern Recognition. New York:
Academic, 1990.
T Detection Accuracy -5.6~-5.7 VF: 84.00%
VT: 95.56%
-5.8~-5.9 VF: 92.00%
VT: 95.56%
-6.0~-6.8 VF: 96.00%
VT: 95.56%
-6.9~-7.5 VF: 96.00%
VT: 93.33%
0 10 20 30 40 50 60 70
-20 -15 -10 -5 0 5
x--VF o--VT
