MAHI Research Project

1
MAHI Research Project

• Guides Dr. D. Medhi
• Dr. Y. Lee
• Members Manish Mehta
• Anup Marwadi
• Hongming Wang
• Quddus Chong

2
MAHI Research Database
The Current System. A Proposed
System. Challenges.
3
Objectives

• Improve the quality of clinical data by clearly
  defining data points and tracking more relevant
  attributes and dynamically add data points.
• Increase reliability, timeliness of data access,
  make it user friendly.
• Provide a means for making Ad-Hoc queries.
• Need for generalization, but in accordance with
  the current hospital standards.
• A means to incorporate wireless standards in the
  existing system.

4
Database Overview

• The database is made up of 3 parts
• Quarantine Area This is the area where all data
  destined for the repository is cleansed.
  Anomalies and exceptions are addressed in this
  section
• Repository Area This is the section where all
  the authorative and clinical data is stored.
  Highly normalized and non-redundant data storage
  area.
• Analysis Area Used for data extraction and
  analyzing the data.

Extract/ Analyze Data
Load/Store Data
Acquire Data
Quarantine
Repository
Analysis
5
Current System Configuration

• Windows NT based.
• Microsoft SQL server.
• MS Visual Basic for the front end. Performance
  critical elements are created using VC. Ad-Hoc
  queries are developed using crystal reports
  provided by VB.
• Front End- The common functions are built using
  COM(adv.) or ActiveX modules. No direct
  manipulation of databases is allowed. All
  accesses through a class object. Secure
  transactions using Commit/Rollback.
• Security Issues!!!

6
Repository Entities

• A patient goes to a location and a caregiver does
  something to that patient. In that process a
  material may be consumed.
• Various tables are studied.
• Relevance of cross-references are yet to be
  determined.

7
Data Mining in MAHI Project

• Data Cleaning
• Data Integration and Transformation
• Data Mining

8
Data Cleaning

• Missing values Ignore the tuple or fill in
  values
• Noisy data Binning, Clustering, and Regression
• Inconsistent values correct manually.

9
Data Integration and transformation

• Data identification problem XY?
• Redundancy problems Correlation analysis
• Value conflict.

10
Data Mining Techniques

• Decision trees Make right decisions
• Association rules Find unknown relationships

11
Data Mining Tools

• Decision trees C5.0
• Association rules reference to SONAR

12
Possible Components

• X_MAKER decision trees
• Y_FINDER association rules
• XY_MIXER decision making based on the \
• association rules

13
MAHI Research Database

• Role of XML technology

14
Benefits
Benefits

• Extensible, reusable, flexible.
• Provide document validation through DTDs and XML
  Schema.
• Existing XML specs for medical systems (e.g.
  SynExML).
• Microsoft provides COM-compatible XML component
  (MSXML).

• Extensible, reusable, flexible.
• Provide document validation through DTDs and XML
  Schema.
• Existing XML specs for medical systems (e.g.
  SynExML).
• Microsoft provides COM-compatible XML component
  (MSXML).

15
XML in Data Movement

• Client input to be formatted as XML for data
  validation and transmission.
• Use of DTDs and XML Schema as validation
  template.
• Capability of XML in performing batch processing.

16
XML in Data Extraction

• Data extracted from data store after query to be
  formatted into XML and transmitted to requesting
  client.
• Modules/components to do XML formatting should be
  part of a middleware layer.
• Web-interface should be designed to allow for
  ad-hoc queries.

17
XML in User Interface

• A dynamic user view can be generated using XSL,
  but should be wrapped by a main user interface
  frame.
• For security/platform neutrality/mobility issues,
  more investigation into XML-related
  specifications like SOAP, WML is needed.
• Two types of client systems to handle different
  functionality
• q Standard query/view used by caregivers.
• q Analysis used by clinical researchers/bio-sta
  tisticians. This client will require a
  customized adaptor to communicate between
  middleware layer and local analysis software
  (e.g. SAS).

18
Data Flow with XML
User query via CGI/ASP/JSP
Quarantine
Web-enabled Browser
Query Generator Component
Data Store
Middleware Server
Basic Client
Customized XSL Stylesheet
Repository
Results (in HTML)
Data Store
Adaptive View Component
Results (in XML)
Adaptor Interface
Analysis
Customized Browser
Results (in XML)
Middleware Layer
Data Store
Research/Analysis Client
19
Tasks to be addressed

• Platform Neutral Architecture
• Mobility of end users
• Simple user Interface
• Security

20
Implementation objectives of my Interest

• Provide ability to move the data in timely and
• efficient manner. ( WLAN )
• Reduce the load on the end terminals as they can
  be
• battery operated devices with comparatively
  low
• computing power (eg. PDA). ( Thin Client )
• Provide security of data and Authentication of
• the users. ( Wireless Security Issues )
• Define different Access Levels for different
  sets of
• users. ( Multi Level Security )

21
Why WLAN ?

• Wireless LAN systems have shown to meet the
  technological and organizational needs of Health
  care market by ..
• speeding diagnostic and case analysis time
  turnaround
• Reducing Hospital Labor
• Reducing Procedural costs
• Reducing Scheduling time

22
Issues ( WLAN)

• Health Affects
• Security
• Cost
• Range
• Throughput
• Interoperability with Wired Networks

23
Thin Client

• The end terminals (can be a wireless device like
  PDA) are to be used to populate the Database by
  gathering the data and sending to the Server for
  processing.
• Similarly the end terminals are to be used only
  to query the data and display the data on the
  users screen.

24
Security and Authentication

• In Wireless Environment we need to get rid of
  unintended receivers ( eavesdroppers ) listening
  to wireless traffic.
• Every user using the system should be
  authenticated by the system.
• Provide Multiple levels of security for different
  sets of users.

25
Issues in Security and Authentication

• Key distribution for encryption of data
• Key management and Key refreshing issues
• Levels of security and their privileges.