Background: Brain atlas-assisted operations, such as targeting in stereotactic and functional neurosurgery, localisation analysis and metanalysis in human brain mapping, or structure segmentation and labelling in neuroradiology, highly depend on the accurate localisation of the landmarks used for atlas-to-data registration. One of practical and widely used registration methods is the Talairach proportional grid system transformation based on the Talairach landmarks. However, there are several problems associated with the original Talairach landmarks. In the Talairach-Tournoux brain atlas, some Talairach landmarks are not available and locations of others contradict their definitions. When dealing with patient-specific data, the definitions of the Talairach landmarks are not constructive enough or make their identification time consuming. Moreover, there is an inconsistency between the Talairach landmarks and Talairach grid.
Method: The modified Talairach landmarks, conceptually equivalent to the original Talairach landmarks, are introduced here. They have several advantages and overcome some limitations of the original Talairach landmarks. Three various intercommissural distances are defined: central, internal, and tangential, and the formulas determining their lengths and errors associated are derived. An efficient method for calculating the modified cortical landmarks is proposed.
Findings: The internal intercommissural distance is the closest to the original Talairach intercommissural distance. Its relative intercommissural error is only 0.5%, as opposed to the central and tangential intercommissural distances resulting in high (about 10%) relative intercommissural errors. On the other hand, the internal and central intercommissural distances result in a high maximum displacement error at cortex amounting to about 11 mm while the tangential intercommissural distance gives only 1 mm error. The sensitivity of the internal intercommissural distance to the actual location of the intercommissural plane is high reaching above 10%. On the other hand, the sensitivity of the central intercommissural distance is below 0.5%. Each of the Talairach cortical landmarks needs three coordinates for its identification that requires the availability of two two-dimensional projections, the generation and analysis of which is computationally expensive. A modified cortical landmark is identified on a single, one-dimensional projection.
Interpretation: The modified Talairach landmarks facilitate the rapid and automated calculation of the Talairach transformation and give more flexibility in their use for specific applications. In stereotactic and functional neurosurgery, the internal intercommissural distance is the most suitable to provide a high accuracy for subcortical structures. In localisation analysis in human brain mapping research, a high accuracy has to be achieved at the cortex and the tangential intercommissural line is superior. Human brain mapping metanalysis may benefit from the use of the central intercommissural distance minimising errors due to the intercommissural plane positioning. In neuroradiology, a high accuracy is required both for subcortical and cortical structures and the tangential intercommissural line should be used.