Imaging informatics tools and techniques are being used in post-analysis or research-oriented studies and effective use of these techniques in clinical applications requires innovations in the following areas:
- Validation of various imaging informatics tools in the context of radiology. This allows the user (radiologist or clinician) to be able to apply such tools to data for extracting insight and decision-enabling information and have certain level of confidence about their performance and reliability. This is the subject of several studies being conducted now.
- Creation of real-time applications of the informatics tools applied to medical diagnostic images. Most of the analytic routines are extremely time-consuming, which precludes them from being used in the daily routines of the radiologists. For example, the process of image registration cannot be applied on demand to images under study, because it may take up to several hours for the results to become available.
- Advanced management of multi-modal medical record content is required to combine the various pieces of multi-modal data and make them available for population and temporal studies, decision support and treatment planning.
To address these requirements, we have developed a multi-modal Medical Analytics Platform, a system that enables real-time collection-level text and image analytics for improved diagnosis and patient care. This system consists of a web client connected to a web server and a FileNet server. The web server is in turn connected to a Blade server housing the IBM Cell Broadband Engine. for high performance computation.
The resulting platform leverages the Cell broadband engine as its high-performance computing platform for carrying out image analytics. Algorithms such as image registration are deployed on the Cell platform. Medical record text analysis is applied to patient records to make them amenable for collection-level search and retrieval based on semantic concepts derived from RadLex taxonomy.
This system is a step towards a better PACS (PACS++) and provides methods for searching radiology reports and clinical notes for patients using standard terminologies such as RadLex, SNOMED CT and ICD-O, and utilizing a high performance compute engine to perform real-time image registration.
The system is used through a thin web client interface in applications such as the TARA demonstration below which utilizes IBM DB2 and Websphere.
Demonstrations
|
|
|
|
Movie of the TARA Text and Radiology Application and the Medical Imaging Informatics
