MiBank: A Web-based Integrated Medical Information System for Traumatic Brain Injury Suisheng Tang 1 , Zhuo Zhang 1 , Boon Chuan Pang 2 , CC Tchoyoson Lim 2 , Beng Ti Ang 2 , Cheng Kiang Lee 2 , Chew Lim Tan 3 , Tianxia Gong 3 , Ruizhe Liu 3 , Qi Tian 1 1 Institute for Infocomm Research, 1 Fusionopolis Way, #21-01 Connexis, Singapore 138632 2 National Neuroscience Institute, 11 Jalan Tan Tock Seng Singapore 308433 3 National University of Singapore, 10 Kent Ridge Crescent, Singapore 119260 Abstract – Approximately ten million people in the world suffer from traumatic brain injury (TBI) each year. A total of $60 billion cost due to TBI was estimated in the United States in year 2000. To reduce the burden more clinical research and education are required. In this study we developed MiBank, a web-based integrated TBI information system, to enable rapid access to both digital images and associated text reports for audit, education and research. MiBank contains more than 30,000 brain computed tomography (CT) images from over 500 patients and is equipped with functional options to search, compare, summarize and annotate CT images, radiology reports and clinician remarks online. The image annotation function is designed to enable clinicians and researchers to capture and display domain expert knowledge, and a discussion forum function encourages active communication and sharing. Emphasizing confidentiality of anonymised data and access control, MiBank provides a virtual collaboration platform integrating various clinical data sets for research and continuing education. As an online information system, it eliminates the restrictions of the traditional isolated DICOM workstations. MiBank can potentially support remote consulting and statistical analysis of aggregated multimodality data. Although MiBank is designed and implemented for TBI, it may be extended and customized to study other clinical disorders. In this report, we share our learning experience through user survey and also propose a future plan to improve the system. MiBank may be accessible by researchers and clinicians on request. Keywords – Traumatic Brain Injury (TBI), Information Integration, DICOM, Image Annotation, Relational Database I. INTRODUCTION The World Health Organization estimates that 10 million people worldwide suffer from traumatic brain injury (TBI) every year, leaving those with severe head injury often dead or severely disabled. Cranial imaging is part of standard evaluation for these head-injured patients [1]. Head computed tomography (CT) is commonly employed to identify fractures, intracranial lesions and mass effect [2, 3]. These imaging findings and other associated information contribute to treatment planning and outcome prognostication [4, 5]. CT images are saved and stored in standard Digital Imaging and Communications in Medicine (DICOM) format, which consists of two parts: 1) a demographic textual component, and 2) a pixel data component. As an industry standard, DICOM has been adopted by vendors to support interoperability among imaging devices. Picture archiving and communication system (PACS) is widely used to manage DICOM files. PACS provides the mechanism for data communication, transmission, storage, and connection to existing hospital information systems (HIS) or radiology information systems (RIS). However, system integration generally lags behind state of the art imaging and storage facilities. In part, this may arise because industrial objectives are often dictated by commercial aims rather than attempts at optimizing the end user experience [2]. In particular, information retrieval from disparate and large data sets in hospitals remains problematic. Often a query of RIS or PACS database is limited by a search for individual patients either by name or identification number [6, 7]. At best, a query returns imaging performed on a particular date or under a general category which includes attending physician and location. Furthermore, DICOM image viewers often do not take full advantage of web-related innovations. Clinically relevant data in patient care, such as physiological parameters, laboratory results, and radiological imaging are often maintained separately. This segregation limits complete data integration on current existing databases. Such segregation is artificial as physicians have to integrate all clinical data in daily decision making. Any ability to extend such integration of patient data to a huge dataset would be a significant advance [1]. In situations where randomized trials are not available or impractical, such integrated data processing ability would further improve the ability for clinical guideline formulation based on best available evidence. Retrospective studies would also have access to complete datasets which parallel those obtained from a prospective collection. Such a system would also allow data from different institutions to be pooled for future studies. In this report, we describe our design and implementation of MiBank, a proof of concept web-based integrated knowledge system for TBI. We also share our experience of conducting a qualitative user survey to align users’ requirements to functional design during system development. II. DATABASE IMPLEMENTATION A. Patient Data Data from severe head injury patients over a 7 year period admitted to the Department of Neurosurgery at the National Neuroscience Institute, Singapore were collated, following 978-1-4244-4764-0/09/$25.00 ©2009 IEEE