Computer, computational data modeling issues

Table 16: Computer/database issues.
Name Description Reference
Computer Cluster fMRI preprocessing with a computer cluster [Erberich et al., 2000]
Peer-to-peer database Napster like service for brain imaging data [Bly et al., 2001b]
QBISM Database for 3D spatial data especially brain mapping data. Built around Starburst DBMS. [Arya et al., 1996a,Arya et al., 1994,Arya et al., 1993,Arya et al., 1996b]
BRAID Object-oriented database augmented with image processing and statistical operations based on IllustraTM [Herskovits, 2000a,Letovsky et al., 1998]
B-SPID Object-relational database for neuroimaging data based on IllustraTM [Diallo et al., 1999a,Diallo et al., 1997a]
BIRN ``Biomedical Informatics Resource Network'': American project to closely link universities for collaboration in neuroimaging by application of SDSC's Storage Resource Broker (SRB).
NeuroCore Database framework used in, e.g., fMRIDC
NeuroML XML for neuroscience, more specifically neuronal modeling [Goddard et al., 2001,Crook et al., 2005]
XNAT Software platform to store MR neuroimaging and related data. The technologies used are Java, DICOM and XML. [Marcus et al., 2007],
-- 3-dimensional database of deep brain functional anatomy for image-guided neurosurgery (IGNS). Coding structure according to Tasker [Finnis et al., 2000,Tasker et al., 1978]

Image Compression

There exist a number of general purpose lossless compression programs, e.g., gzip and compress. These programs seldom make a very efficient compression of neuroimaging data. SmallTime [Cohen, 2000], ( is a program specifically designed for MRI sequences and works with differences of 16 bit images. Other algorithms have been described [Thiran et al., 1996,Adamson, 2002,Wu and Forchhammer, 2004], some based on wavelets.

The DICOM standard specifies lossless compression via the JPEG-LS (ISO 14495-1) and JPEG 2000 (15444-1) standards.


Most databases management systems (DBMS) are relational. These consist of tables and links between tables. A wide-spread query language is SQL. Early relation database management systems (RDBMS) had very simple types: integer, floats, string, etc. and would not represent 3D neuroimaging data well. Extensible DBMS such as so-called object-relation DBMS (ORDBMS) augment the ordinary relation DBMS with (complex) user-defined types (objects). One of the first ORDBMS was POSTGRES, see, e.g., [Stonebraker and Kemnitz, 1991]. This database was commercialized by Illustra, which in December 1995 was bought by Informix. Informix was in turn bought by IBM in April 2001. Another extensible DBMS is/was the Starburst DBMS, see, e.g., [Lohman et al., 1991]. It's query language ``hydrogen'' is based on SQL. Open source ORDBMS version exists with PostgreSQL ( EXODUS is yet another extensible database system (Carey et al, 1991).

Image databases

Content-based image retrieval (CBRI) uses image feature, e.g., for finding similar images in a database.

One finds descriptions of general systems (not necessarily for neuroscience images) with Photobook [Pentland et al., 1993],, QBIC [Flickner et al., 1995] and Virage [Bach et al., 1996].

Medical image databases

[Tagare et al., 1997] discusses some of the issues in content-based retrieval approaches in medical database.

Clinical medical image databases are often refered to as ``picture archiving and communication systems'' PACS [Huang and Taira, 1992]. There are probably many commercial systems in this area. One such is EasyViz by Medical Insight.

Neuroimaging image databases

[Bjaalie, 2002] discusses neuroscientific databases, particularly the NeSys database, and [Kötter, 2002] is an edited book about neuroscience databases. Other methods of image retrieval are presented in [Liu and Dellaert, 1998a,Liu and Dellaert, 1998b,Liu et al., 1998,Liu et al., 2001b,Liu et al., 2002].

A review of databases for neuroimaging appears in [Diallo et al., 1999b,Diallo et al., 1997b].

Image archives and data sharing in research oriented brain mapping have been discussed in a number of articles [Chicurel, 2000,Aldhous, 2000,Marshall, 2000,Nature editorial, 2000,Cohen et al., 2001,Barinaga, 2003,Toga, 2002b]. This discussion was initiated after suggestion for required submission to the fMRI Data center [Van Horn et al., 2001a,Van Horn and Gazzaniga, 2002]. A review of neuroscience atlas/databases is [Van Essen, 2002].

[Wang et al., 2006] implemented content-based image retrieval on fMRI contrast maps with a combination of wavelets and image feature extraction. [Cho, 2005] discusses a retrieval system with fMRI and single unit recordings. The CCVT tool also enables query-by-example [Cornea, 2005]. NeuroServ and NIRV are software tools mentioned by [Carley-Spencer et al., 2006]. Table 17 lists other content-based image retrieval systems in the neuroimaging context.

Table 17: Medical image retrieval systems
Name Area Features and description Reference
-- Lymphoproliferative disorders Shape (Fourier descriptors), texture (MRSAR), color. Image-based queries, trained classifier [Comaniciu et al., 1999,Comaniciu et al., 1998].
-- (Synthetic) MRI Topographical relations, size, roundness, orientation. Images represented as ``attributed relational graphs'' [Petrakis and Faloutsos, 1997,Petrakis, 2002]
Brede Database Functional neuroimaging Finding nearest Talairach coordinate to a given location. ``Grey-level'' volume from voxelized locations. Determination of related volumes in the Brede Database. [Nielsen and Hansen, 2004b,Nielsen, 2003,Nielsen and Hansen, 2002b],

Dermatlas Clinical dermatological images (photographs, histological, biopsy) Text-based (keywords, diagnosis, body site, pigmentation, ...). Internet-based with collaboration (users are able to submit new images).
I$ ^2$Cnet MRI, ... ROI (location, shape, size: roundness, compactness, area, orientation), texture (maximum probability, angular second moment, contrast, inverse difference moment, entropy, correlation, variance, cluster shade, diagonal moment, k statistics, fractal signature). Internet-based. [Orphanoudakis et al., 1996]
(Jerne, ``Related volumes'') Functional neuroimaging ``Grey-level'' volume: Voxelized locations. Determination of related volumes in the BrainMapTM database by voxelization [Nielsen and Hansen, 2004b,Nielsen and Hansen, 2002b],
(KMeD) MRI, x-ray, ... Shape, size, texture, topological relations. Spatial temporal query language: (KSTL). Features in a hierarchical structure [Chu et al., 1998]
MIMS   Image descriptors: file type (JPEG, GIF), device (MRI, radio), domain (radiology), ... Topographical relations. Java Internet-based. Stores images voice reports and general (text) data. Thesaurus on descriptors [Chbeir et al., 1999,Chbeir et al., 2000]
WebMIRS X-rays
xBrain Functional neuroimaging Search for nearest Talairach coordinates.

Neuroinformatics database

In neuroimaging contexts the neuroimages themselves do not only have to be databased but all the associated data as well. These associated data might be subject information such as demographic, clinical and genetic. Complications for the construction of databases for such data might be extensibility and the secure transportation of data across multiple sites. Thoughts on the problems for such systems appear, e.g., in [Bockholt et al., 2006] (``The MIND Institute'') and [Keator et al., 2006] (``XCEDE'', ``BIRN'' ). Some form of relational database for neuroscience data is described by [Rudowsky et al., 2004].

Unique identifiers for authors are partially implemented in the Brede Database [Nielsen, 2003]. This has also been suggested by [Falagas, 2006] for more broader application.

Computational neuroscience

The Blue Brain Project aims at modeling the neocortical column of the somatosensory cortex of young rats based on the NEURON program and using supercomputers, and is introduced in [Markram, 2006].


OSIRIS [Ligier et al., 1994]: This seems to be an image analysis/processing program rather than a database.

M. Bota and M. A. Arbib. The NeuroHomology Database. Neurocomputing, 38-40, pp. 1627-1631, 2001.

[Hirsch and Koslow, 1999]

[Miller et al., 2001,Gardner et al., 2001b].

[Guimond et al., 1997]: retrieval of corresponding brain structures from a database of medical images.

Database Challenges and Solutions in Neuroscientific Applications (1997). Ali E. Dashti, Shahram Ghandeharizadeh, James Stone, Larry W. Swanson, Richard H. Thompson

Finn Årup Nielsen 2010-04-23