Medical Text and Image Analytics       

links

 Anni R. Coden photo photo

Medical Text and Image Analytics - overview


Medical Text Analytics

The Medical Text Analysis System (MedTAS) is a UIMA-based, modular and flexible system that uses advanced Natural Language Processing (NLP) techniques to extract structured information from unstructured data sources, such as pathology reports, clinical notes, discharge summaries, and medical literature. MedTAS is also a development platform that is adaptable to customer and domain requirements. It has been designed to operate within institutional systems, and seamlessly integrates with IBM products, such as the DB2 Data warehouse. Implementations of MedTAS have been developed for major medical institutions. MedTAS/P is a version customized for the pathology domain. It is based on a novel representation of cancer, its characteristics and disease progression.

High Performance Medical Imaging

Our group has a focus on high-performance medical image analytics. We have successfully demonstrated the power of the IBM Cell broadband engine in this space, gaining up to 60x speed-up on a linear registration task using cell over an x86 implementation. We are extending the registration algorithm for soft tissues using a non-linear transformation. Another effort is in the area of benchmarking – comparing the performance of key image analytics algorithms and their sub components between implementations on the IBM Cell Broadband Engine and those on GPU implementations. Of particular interest are CT Reconstruction algorithms.

Multimodal Medical Informatics

The Multimodal Medical Informatics project aims at providing systems and tools to support clinicians' decisions and reduce uncertainties at different stages of the clinical care by: analyzing, organizing, fusing and summarizing information embedded in patients' multimodal medical records for enhanced access to data. We built a prototype – the Medical Analytics Platform geared towards helping Radiologists access and process images taking advantage of the IBM Cell Broadband Engine and correlating them with analyzed and relevant textual sources such as pathology and radiology reports.

Some of our Partners

  1. The Mayo Clinic
  2. Mayo-IBM Medical Imaging Research
  3. Swansea University
  4. IBM - Memorial Sloan Kettering partnership
  5. IBM - New Jersey Cancer Center partnership

Contacts

  1. Anni Coden
  2. Chalapathy Neti