Background
Tractography has been used in various studies with respect to the improvement of patient-specific DBS targeting. Nevertheless, methodological influences of the chosen parameters and associated errors are often neglected. The aim of this study was to estimate concrete errors associated with specific image processing steps when using measurements of distances to specific subcortical fiber tracts to predict optimal stimulation sites for DBS targeting.
Method
Probabilistic tractography of the crossing and non-decussating part of the dentato-rubro-thalamic-tract (c-/ nd-DRTT) was performed using FSL 6.0.3 in 40 PD- and ET-patients having received bilateral DBS surgery. DBS-electrodes were reconstructed using LeadDBS. The influence of (1) the choice of threshold for binarization of fiber tracts, (2) manual measurements compared to measurements using automized distance maps and (3) normalization into the MNI standard space on measured distances were investigated.
Results
Different thresholds for binarization resulted in non-linear and unpredictable variations of measured distances up to 1.72 ± 1.49 mm (mean value ± standard deviation). Manual measurements on the axial slice of the electrode contact showed a mean error of 0.91 ± 1.36 mm (maximum 14.9 mm) compared to automated measurements. Regarding normalization, a mean error of 0.82 ± 0.50 mm (maximum 2.34 mm) was found compared to measurements in native space.
Conclusion
Measured maximum errors reach up to several millimeters, which might have significant impact on clinical targeting in DBS. Researchers should be aware of these errors and define individual standards for specific studies.