Imaging Informatics-based Multimedia ePR System ...

Imaging Informatics-based Multimedia ePR System for Data Management and Decision Support in Rehabilitation Research Ximing Wanga, Sneha Vermaa, Yi Qina, Josh Sterlinga, Alyssa Zhoua, Jeffrey Zhanga, Clarisa Martinezb, Narissa Casebeerb, Hyunwook Kohb, Carolee Winsteinb, Brent Liua a

Image Processing and Informatics Lab, University of Southern California, 734 W Adams Blvd, Los Angeles, CA 90089; b Motor Behavior and Neurorehabilitation Lab, Div. of Biokinesiology and Physical Therapy, USC; ABSTRACT With the rapid development of science and technology, large-scale rehabilitation centers and clinical rehabilitation trials usually involve significant volumes of multimedia data. Due to the global aging crisis, millions of new patients with age-related chronic diseases will produce huge amounts of data and contribute to soaring costs of medical care. Hence, a solution for effective data management and decision support will significantly reduce the expenditure and finally improve the patient life quality. Inspired from the concept of the electronic patient record (ePR), we developed a prototype system for the field of rehabilitation engineering. The system is subject or patient-oriented and customized for specific projects. The system components include data entry modules, multimedia data presentation and data retrieval. To process the multimedia data, the system includes a DICOM viewer with annotation tools and video/audio player. The system also serves as a platform for integrating decision-support tools and data mining tools. Based on the prototype system design, we developed two specific applications: 1) DOSE (a phase 1 randomized clinical trial to determine the optimal dose of therapy for rehabilitation of the arm and hand after stroke.); and 2) NEXUS project from the Rehabilitation Engineering Research Center(RERC, a NIDRR funded Rehabilitation Engineering Research Center). Currently, the system is being evaluated in the context of the DOSE trial with a projected enrollment of 60 participants over 5 years, and will be evaluated by the NEXUS project with 30 subjects. By applying the ePR concept, we developed a system in order to improve the current research workflow, reduce the cost of managing data, and provide a platform for the rapid development of future decision-support tools. Keywords: Rehabilitation engineering, electronic Medical Record, ePR, System integration

1. INTRODUCTION With the rapid development of science and technology, the present-day rehabilitation research and training approaches are equipped with modern technology, such as computer-aided data-processing, statistical methods, database systems, nanotechnology and medical imaging. Enhanced with the new technology, large-scale rehabilitation centers and clinical rehabilitation trials usually involve a large volume of treatment recording data, questionnaires and multimedia data. The multimedia data are in a variety of forms, including biometric waveform data (EMG kinetic, etc.), video clips, force vectors data and medical imaging data, such as MRI and TMS(Transcranial magnetic stimulation).Moreover, due to the global aging crisis faced by today’s world health care system (for example, U.S.’s 76 million baby boomers begin to retire), millions of new patients with age-related chronic diseases (such as Chronic Spinal Cord Injury, hip fracture etc.) will produce huge amounts of data and contribute to soaring costs of medical care. In addition, data processing and analysis is hampered due to the lack of a platform infrastructure for analyzing data. Collaboration between different sites is also difficult to conduct because of the lack of tools for data sharing. In rehabilitation engineering research centers, as well as clinical rehabilitation trials, effective data management must address several challenges: tracking each participant through the workflow, collecting related data, efficiently sharing data to facilitate collaboration in different sites and providing a platform to develop a tailored computational and predictive statistical model for decision support for an individual subject. Therefore, we aim to extend the concept of electronic patient record (ePR) system to the field of rehabilitation. The system also provides a platform for the development of data processing tools, such as statistical modeling and computer aided detectiion/diagnosis. As an initial first start, we developed tw wo specific sysstems based on imaging infformatics concepts for the NEXUS prroject and the DOSE D trial as examples. e Medical Imaging 2013: Advanced PACS-based Imaging Informatics and Therapeutic Applications, edited by Maria Y. Law, William W. Boonn, Proc. of SPIE Vol. 8674, 86740P © 2013 SPIE · CCC code: 1605-7422/13/$18 · doi: 10.1117/12.2008585 Proc. of SPIE Vol. 8674 86740P-1 Downloaded From: http://proceedings.spiedigitallibrary.org/ on 07/12/2014 Terms of Use: http://spiedl.org/terms

2. ME ETHODOLO OGY m, we analyzed d the workflow w of specific studies using database and data model sstandards To realizze the system (DICOM, HL7 and IHE) and developeed a subject or patient-oriennted system w with basic feattures and com mponents. Furthermore,, the system is customized baased on the specific requiremeents of the projject. velopment 2.1. Inffrastructure dev To impllement the sysstem , we deveeloped a proto otype system an and customizedd them based oon the requirem ments of specific projeects. As shown n in the figure 1, the basic sy ystem has a cenntralized databbase and a file storage system m, a webbased graphiccal user interfaace and platform ms for integratting decision-suupport tools annd data miningg tools. The excchanging multimedia data d include su ubject text dem mographics dataa, treatment reccords, screeninng and evaluatiion data, imaging data, video, audio,, EMG, wavefo orms and otherr formats of thee multimedia ddata. hical

41-

Interface

Treatm Recor

Screeni &Evaluat

Ima in

I I

Video,au io,E

cision-

ort Too

1

I

Il1..

.-11.

)ata base

MG,wave orm

CnJ vv1oh_Rn G 1.I LO GI JCW

1A/ß

,

eP'R sy51te m ------

,' "

Figure 1 logic model m of the prototype multimeddia ePR system ffor rehabilitationn

m, there are several features need to be coonsidered baseed on the geneeral rehabilitattion trial To realiize the system workflow. Data entrry 2.1.1 c demographics/quesstionnaire dataa. • E-forms for collecting nd displaying multimedia m datta(videos, audioo, kinetic, forcce vector etc). • Uploading an dy anonymizatiion and upload ding. The imagges uploaded w will be automattically anonym mized and • Imaging stud linked by sub bject ID. 2 Data pressentation 2.1.2 • Traditionally y database-baseed retrieval.

Proc. of SPIE Vol. 8674 86740P-2 Downloaded From: http://proceedings.spiedigitallibrary.org/ on 07/12/2014 Terms of Use: http://spiedl.org/terms

• •

A web-based zero-footprint DICOM(Digital Imaging and Communications in Medicine) viewer with annotation, ROI(Region of interest) and measuring tools. Provide a platform for integrating image-processing modules and data mining/computational models. We aim to develop an online tool for applying processing methods and displaying the processing result in the system, which will support the decision-making.

2.1.3 Data sharing. The system also includes a secured sub-system for sharing data between users in different physical sites. • Security. Compared to public data sharing systems (e.g. Dropbox, MS Sharepoint or Google Drive) which do not meet standards for protecting participant confidentiality, this data sharing sub-system allow user to store the data in trusted private data center (which has more privacy), • Simplicity. Although public data sharing systems have lots of advanced features, sometimes they are too complex to learn and the customization is difficult. This light-weight sub-system is straightforward and easy to customize based on specific needs. 2.1.4 User control. • User groups. The system has the feature for granting different user rights to different user groups. • User activity tracking and logging. All users’ activities will be logged, which is essential for conducting research. 2.1.5 Client side access. • The system supports access from desktops, laptops, and mobility solutions (eg, iPAD, iPhone). 2.2 Customized system based on DOSE workflow Based on the model described in the section 2.1, the system was designed for a clinical trial called DOSE and the system was implemented a as an example. DOSE trial aims to determine the optimal dose of stroke rehabilitation for the arm and hand, with a target subject recruitment of 60 stroke subjects with imaging and data collection. As shown in the figure 2,each subject will go through three stages: enrollment, participation, follow-up. In the enrollment stage, each subject will be screened using a series of behavioral assessments. The referred subject will be screened in person and randomized in to different groups. If the subject is enrolled, in the participation stage, the subject will go through three week-bout therapies and the multimedia treatment data, evaluation data, and medical imaging data will be collected. In each therapy cycle, a subject needs to take a pre-test bout, therapy session and a post-test bout. At the follow-up stage, each subject needs 6 monthly follow up. In each follow up session, evaluation data is collected. Upon study completion, the data will be analyzed.


In- Person

D()SE referral

itial EValíBaSe line

R nndomization

Eligibility

Test 1

Screening

Post Test Hal Eval/ Batseline 1

Initial Eval I Baseline

IBout

1

Pre -

Test

' 1Bout 2

(Bout 1

I

I

r r T

IBout 31713

>

Weel k Bout The 'apy 2

Veek Bout Therapy 1

Week Bc Therapy

1 nonth

low up -

2 month Follow up

3 non th Follow

up

*

4 month

5 month

Fc)Ilow up

Follow up

6 ilio nth Follov vup

Figure 2 Workfflow for each sub bject in the dosee optimization reehabilitation triall

o the flow, thee system is cusstomized to three sections, s creening/enrolllment, treatmeent and evaluattion, and Based on report data an nalysis. The sccreening compo onent not only collects testingg data, but alsoo automaticallyy generates sugggestions for making decisions d on en nrolment based d on specified rules. r In the trreatment and evvaluation part,, the system prrovides a data entry solution for diifferent types of data, and also displays the status of the trial’s progress for eacch study participant. The T report dataa analysis part aims to provid de not only datta retrieval, buut also general statistical toolls. These tools will sh how the analyssis results onliine once the data d is collecteed. Therefore,, users can avoid a tedious work of downloading g data and analy yzing them. ve user groups in i the system. There are fiv • Administrato or user group. The administrrators are respponsible for thhe managing thhe system andd regular users and baacking up dataabase regularly y. The adminiistrator is assiigned to a thirrd party who will not participate in n the study. ment data and any data • Evaluator user group. The study requiress the evaluatorrs to be blindeed to the treatm void any bias over o the evaluaation. analysis to av • Physical therrapist user grou up. Physical th herapists are reequired to be bblinded to anyy evaluation daata. They only have acccess to the treaatment recordin ng module. • PI/inspector user group. PII or inspector have the abiliity to see the whole trial daata and status,, but not m any dataa. allowed to modify • Researcher user u group. Ressearchers are ab ble to downloaad the data andd do the data annalysis, but nott allowed to modify or enter any data. The useer control mecchanism provid des a full logging of all userrs’ action for ssecurity reasonns and gives eeach user specific rightts based on their responsibilitty. a a web server,, with a databaase and a file sttorage system. A web based ggraphical user interface The sysstem is set up at is developed d for users to manage the system. s Based d on the databbase and file storage system m, we aim to develop computationaal models and decision-supp port tools. Currrently the com mputational moodels and deciision-support ttools are still in progrress. The multtimedia data used u in the sy ystem includess the patient/suubject text deemographics, trreatment records, screeening and evaaluation, MRI studies and TMS T studies daata. Client acccess methods iinclude PC, M MAC and mobile devicces such as iPA AD, smart phon nes.


2.3 Customized system m based on NE EXUS workflow w In addittion to the DOS SE trial, the sy ystem was also designed baseed on a secondd project calledd NEXUS. NEX XUS is a project at Reehabilitation Engineering E Reesearch Centerr (RERC), US C, which aims to assess thee usability of standing balance virtu ual reality gam mes specifically y designed forr rehabilitationn. These gamess are designedd to engage thee user in therapy exercises in a con ntrolled and cu ustomized man nner. Figure 3 shows the floow of the NEX XUS study. Onnce each bject will take a pre-interview w session, setuup session andd participate in games and thhe data is subject is enrrolled, the sub acquired afteer the game. A post-interview p w session will be done to concclude the cycle. For the data ccollection, a paarticipant in a single-seession counterrbalanced two different dynaamic balance taasks in which the participannt reached for a virtual target ( virtu ually projected d jewel) or a real target( tennis ball), maatched for the number and position of taarget and movements under u an orderred set of initiaal conditions(sttatic standing; dynamic steppping). The oveerall session inncludes a set of questio onnaires aboutt health of the participant, prrior experiencee with technoloogy and video games, and hhistory of falls. Table 1 shows all the data related wiith each stages.

C:nrollment

F 're

Intervie

Gam e: Reach ing

W

Setup

Non Ga me: Reachi ng

(nmt

NUnn glIVII (tea ,.8,4 n,c

I

Steppi ng

Steppi ng

ME easurements:

HiBart

Post lotarview

rate et c.

Figure 3 Work kflow of the NEX XUS trial

Workfflow Stage PreInterview

Clinical Form ms

A Acquired Dataa

• • •

• • • • •

•

Setup

• • • • • • • •

Informed consent c Demograph hic Information n Activities Specific Baalance Confidencee Scale Immersive dencies Tend Questionnaaire Tellegen Absorption A Scalle Digit Symb bol Coding Ruff 2 & 7 Test Four Squarre Step Test Berg Balan nce Scale EMG Electrode Placemen nt Force Platee Placement EMG Max x Muscle Forcee Test

Immersive Tendency to V Virtual Reality Balance Dynamic B Demographhic Informationn Cognitive P Processing andd Memory Attention

N None


• • Data Collection

Post Interview Data

• •

Depth Set-up Accelerometer & Heart Rate Meter GAME: Reaching GAME: Stepping

• •

NON-GAME: Reaching NON-GAME: Stepping

•

Accelerometer & Heart Rate Meter

• • • • •

Game Engagement Form Slater-Usoh-Steed Form Participant Likert Form Interview Protocol Final Question

• • • • •

Probe Reaction Notes Kinect Data Video Data Accelerometer Data Probe Data

• • • • • • •

Probe Reaction Notes Kinect Data Video Data Accelerometer Data Probe Data Heart Rate Notes

• •

Post Exposure Questionnaire Post Exposure Interview

Table 1 Data to be collected in each stage of the Nexus workflow

The NEXUS project includes data collection stage and data processing stage. For the assessment of multimedia of data which is collected during the data collection process, efficient integration of various data objects would be beneficial to assist in data mining and knowledge discovery, which can give access to various clinicians towards the data in efficient and robust way. We developed one such system, which stores video files of participants doing all tasks, audio of participants giving post interview, text based forms as well as raw data from sources such as Kinect, accelerometer together with the basic data such as heart rate. To facilitate the data sharing and post-processing, the system focus on providing a solution for distribution of raw data, capture post-processed information, and query retrieval. The raw multi-media data include DICOM medical images, video files and audio files. The DICOM viewer of the system will enable user to look through the images and mark ROI on the images. Moreover, post processing results, such as CAD in medical imaging will be also stored in the database. A data retrieval based on CAD will also be used to facilitate the research. Currently, the data that is being postprocessed by different assigned teams working on each of the component of the data, who find the relevance based on many factors. This system will be used as a platform for their collaboration and knowledge discovery.

3. RESULT 3.1. DOSE system We have implemented the platform for the DOSE system. The computational models and decision support tools are still in the design. Figure 4 is the personal page for each subject. Once a subject is enrolled, this system will show a summary page to manage all the subject-related data. A timeline overview of participant’s trial shows the status of the subject’s treatment progress. This overview shows the subjects progress in each stage to facilitate researchers study. Below the timeline overview, users can access the data entry part in each stage of the study for this subject.


El] pmau

dase.usc.edulsludy_subjemphp?SludylD=xra,

6

a

California

Patient Informati

Study Flowchart

Timeline Overview of Participant' s Trial Status

®y

®

®®

:11

Initial Eval/Baseline 1

Required Digital Forms for Particular Trial Stage

Mota AcIrnly LagiM.ua1

o.upm.P.,mon Tel MUFTI

Baseline 2 ourttm Us. Twit

MotuawPVipStwua

sean.ewaaaSxwryL !if rugi aore..ao, Component

r,wtaPsnnTattwuPn Orientation

-

55W Fm ut ASAP PaKO.wo agreement tuen

ASAP Tat, wnaEp..u. and PnpnrT..Tt.pLW SOAP Took Cal

Utol aula Pos.iiy Ta. Threshold 13

Rating Fern

Figure 4 Screen shot of the subject personal page for DOSE system

Figure 5 is an integrated DICOM viewer with measurement tools, ROI tools and annotation tools. The screen shot shows a measurement. The zero-footprint DICOM viewer is an integrated component for data presentation.


a;e;ouuy lanal ntopuiNt ION ez-to-zloz - (bw) **MS pww -0zfn3,1100001 10o00/) .MOAevf.aw : ro01.aW%-,an,ag ,..vag .P.M

(oz

.

Idl=epsapnlsNlwpe/MnuyopeAynpa.osn-esop

im to 4

0.

Figure 5 Screenshot of the DICOM viewer with a measurement

The system is currently used by the DOSE trial team including three user groups. 10 subjects have been enrolled in the trial, and 60 subjects enrollment are expected in the future 3 years. The screening and part of evaluation data of the enrolled subjects have been entered into the system. Therapy data is still in the progress of collecting. To evaluate the system, we are collecting user’s feedback through questionnaires and surveys. Table 2 shows areas of issues we will be focusing. Time improvements will be compared w/ current paper-based workflow. First we will evaluate the time improvement between using the system and using traditional paper-based research. The time evaluation session include screening of a subject, enrollment decision, evaluation of a trial subject, weekly bout therapy, follow-up evaluation, report generation, and retrieving data. Additionally, we are going to evaluate the cost of data storage, reliability of the system, user’s suggestion and feedback by questionnaire, system performance, data recovery, user’s satisfaction and data security.


The time improvement of

Other issues

Screening of a subject

The cost of data storage

Enrollment decision

Reliability of the system

Evaluation of a trial subject

User’s suggestion and feedback

Weekly Bout Therapy

System performance

Follow-Up Evaluation

Data recovery

Report Generation

Questionnaire for User satisfaction

Retrieving and query the data

Data security

Table 2 Issues for evaluating the DOSE system

3.2 NEXUS system The nexus system has been implemented and the data of 60 subjects are collected in the system. The data include demographics data, EMG probe data, kinect data, video data, acceleration data and audio interview data. Figure 6 shows a screen shot of the data search. The search criteria include age, gender, ID, marital status, Hispanic/Latino etc. Before the system is developed, the NEXUS trial encountered difficulties in searching for data based on different specifications. The search function provides a method to query data based multiple specifications of patient. For example, the system allows user to query all the subjects between a specific age range. Users are also able to search new conditions among the results from the previous search. Found 5 results

Search by Age TO

56

ID

Demographics

0201010010ÁJ

+

65

EMG Probe Data +

Kinect Data

Video Data

Acceleration Data

e

reach ing_nongame. min ste pping_nongame. min

stepping_game.min reaching_game.min

0201010018ES 0201010021YV 0201010032FS 0201010039TG

Figure 6 Screenshot of data query in the NEXUS system


Audio Interview Data

The figure 7 shows a web-based video viewer for viewing video files. If a video or audio file is found, then the viewer allows users to view it online to decide if he wants to download it. Create:Delete

lin