Martin Rausch, Novartis Pharma AG, Central Technologies, Basel
www.biomap.ch

Anaysis of dynamic contrast enhanced MRI data is an example of BioMAP's capabilities: The original data is shown in the upper image. The Model-Fit-Tool was used for fitting the selected model-function to the temporal profile of one voxel. The lower image shows the result of the fit for the entire image.

In-vivo imaging of compound effects in animal models or humans is of increasing importance for the development and optimization of new drugs. Readouts are usually based on one of two general strategies: Structural analysis is employed, e.g., for the quantification of lesion size and can be supported by tools allowing efficient definition of ROIs or automated segmentation of image data. Secondly, techniques acquiring multi-parametric or dynamic imaging data have been developed that allow assessment of physiological parameters, which are closely linked to the disease process or the effect of a drug. Their quantification requires complex physiological modeling and a variety of mathematical tools, which are usually not integrated in standard software packages provided for image acquisition or scanner control.

With BioMAP it was intended to setup an image analysis platform that can be used for a large range of imaging modalities, and which allows efficient and flexible modeling and quantification of imaging data in pharmacological research and development.

BioMAP was completly written in IDL. It provides a common visualization and storage platform, which can be used for visualization of data from any source, provided that an import filter exists for this format. A variaty of applications can be tailored on this platform by combing various IDL-based software compounds, individually designed for analysis of specific data sets. Visualization is based on multi-planar reconstruction allowing extraction of arbitrary slices from a 3D-volume. Other features linked to visualization are overlaying of two individual data sets or displaying ROIs. BioMAP uses a 4-dimensional representation of image data throughout. The first three dimensions are used for the position of a voxel in space, the fourth dimension is reserved for independent parameters such as time, mass, wavelength or diffusion weighting. In addition to the voxel data, a header is used to store information for scan identification, the physical position of the object in space (important for coregistration of multiple scans) or the imaging protocol. Image data can be imported from several sources (see table below). Processed data, however, will always be stored in Analyze-format. This format was implemented since it is widely used and can be loaded by several other tools, like ANALYZE (Mayo clinic) or SPM (Welcome department of cognitive neurology, London).

BioMap comprises all tools, which are required to process single-subject-data, to combine results from several subjects or sessions and to document the final result of a study. Import and export of partially processed data to other software-tools can therefore be avoided.

The following table gives a functional summary of the current version of Biomap and demonstrates the flexibility and power of this IDL-based medical application. Additional data input or analysis routines can easily be added to this software thanks to its open architecture.

Data Import

• Bruker
• Dicom
• Analyze
• MALDI (Mass spectroscopy imaging, in-house format)
• NIRF (Near infrared imaging, NRC-format)
• Pixmaps (Tif, …) for histological data

MRI-Analysis

• Perfusion-weighted imaging (Bolus-tracking-Method)
• rCBV
• TTP
• MTT
• MSD
• g-Variate-Fit
• FWHM
• CBVrel, MTTrel
• Flowquantification
• Permeability/perfusion-separation
• Permeability imaging (Dynamic-contrast-enhancement-method)
• Initial-slope
• Two-compartment-model
• Diffusion weighted imaging
• ADC
• Trace
• Anisotropy

General Analysis

• Histogram analysis and classification
• Cluster analysis
• ROI-analysis (Single subject ROI readout multi-subject ROI averaging)
• ROI-plotting (Single subject and multi-subject trace averaging)
• Singular-value-decomposition
• Basic calculations (subtraction, scaling, …)

Fitting and filtering

• Fitting of mono-exponential functions
• Baseline correction
• Spatial and temporal filtering

Geometrical operations

• Co-registration of scans of any source, e.g., MRI to Histology, MALDI to Histology, MALDI to MRI
• Elastic matching (removal of motion in dynamic scans)
• Reslicing of image data to a common geometrical space
• Reordering