gms | German Medical Science

For optimal diagnosis in the clinical routine, an accurate comparison of the various image data is urgently needed [Ref. 1]. To even detect minimal morphological variations, the slice images should have the same position and orientation. This adjustment is either manually using visual comparison with the initial data, or automated with a retrospective registration. These methods are very time-consuming and the necessary data interpolation may cause artifacts. In longitudinal MRI exams, a prospective registration can be used to correct misalignments by adaptation of the gradients without the need for patient repositioning [Ref. 1], [Ref. 2]. We implemented the presented method in the image reconstruction system of the tomograph, using the open-source library „Insight Segmentation and Registration Toolkit“ (ITK) developed by the National Library of Medicine (http://www.itk.org).

The procedere presented here is a further development of the work described by Gedat et al. [Ref. 2] for prospective registration of MR data of the head based on the reduced, three-dimensional information of the scout series. Disadvantages of the original solution was the external processing and the resulting lack of automation. The purpose of this work was the implementation of a customizable registration procedure into the image reconstruction and the integration of the slice position correction within the measurement sequences of a modern MR imaging system. For this purpose an interface to the cross-platform C++-library ITK was created.

This method was implemented on a Siemens 3 Tesla MAGNETOM Trio with the local software environment IDEA VA25A for NUMARIS/4 syngo MR 2004A. The basic gradient echo sequence (GRE) has been adjusted as localizer and with a 280mm FOV, 256 x 256 matrix and 10mm slice thickness, three two-dimensional orthogonal images are acquired as standard scout series.

To test the developed method in a practical context, the examination of the effects of ethanol on the changes in BOLD-imaging fifteen healthy and right handed volunteers were examined after written consent according to the local ethics committee [Ref. 3]. A test/re-test design was used to examine the effects of ethanol administration in each individual, i.e., all volunteers were examined before and after ingestion of alcohol. First and foremost the subjects were in sober state and motor skills should be investigated. In the course of the experiment the subjects got a weight-adopted amount of alcohol to drink. The repetition of the tasks were the base for the investigation of the changes in the BOLD signal. As the volunteers had to leave the scanner for ethanol administration, the test setup consisted of two series of measurements with different positions as expected. Because the two studies took place in quick succession, the volunteers took each time a relatively similar position and differed only slightly. For an accurate comparison, the second examination with alcohol was prospective registrated with the initial examination using the developed procedure.

In the scout images of the follow-up examination rotation angles of -3.7° to 2.8° and translation parameters of -10.1 mm to 5.4 mm was detected. This could be reproduced in a retrospective registration with absolute deviations from Δφ= 0.03° ± 0.37° and Δt = 0.44 mm ± 0.51 mm by SPM5 (http://www.fil.ion.ucl.ac.uk/spm).

An vasoactive influence of ethanol on the neurovascular coupling and thereby to changes in the BOLD signal could be confirmed. Between the initial data and those applied with ethanol, significant differences in hemodynamic response function were identified. The developed method for automated prospective registration was successfully tested in a practical context. The complete automation of the application reduce error-proneness as well as time-exposure. With the interface to ITK a very powerful tool was implemented. The large number of integrated algorithms, particularly the various similarity metrics, allow a very rapid adaptation of the method to other areas of application. Due to the further development of the library a constant updating of the used algorithms are guaranteed.

The developed technology allow fast and reliable realignment of the initial slice position and orientation in longitudinal MRI exams without patient repositioning. The prospective registration used only the reduced three-dimensional information of the localizer, the three orthogonal slices acquired with each MR examination. One of these measurements takes less than 10 seconds. This allows the adjustment of the geometrical parameters of the follow-up examination without loss of time in clinical routine. The interface to the ITK library enables the use of a variety of interpolation methods, image metrics and optimization algorithms. That gives us the opportunity for a simple adaptation of the application. Conceivable would be the extension with additional segmentation algorithms or a multi resolution registration.

We would like to thank Dr. Egbert Gedat for the provision of the phantom.