Intra-operative Brain Shift Correction with Weighted Locally Linear Correlations of 3DUS and MRI

Evaluation of MRI to Ultrasound Registration Methods for Brain Shift Correction: The CuRIOUS2018 Challenge

“…Multimodal deformable registration between the MRI and intraoperative 3DUS was achieved with a weighted version of the locally linear correlation metric (LC 2 ), correlating MRI intensities and gradients with ultrasound, while adapting both hyper-echoic and hypo-echoic regions within the cortex. The method [15] was initialized with a global rotation of the US volume to match the orientation observed on the MRI. This was achieved using a PCA of the extracted inferior skull region, identifying the principal orientation vectors of the head, followed by a scaling and translation correction.…”

“…The distances between homologous anatomical landmarks between iUS and MRI were used to assess and rank the registration quality. The CuRIOUS2018 challenge received 8 initial submissions [10][11][12][13][14][15][16][17]. Seven teams validated their methods on the testing data, and six participated in the final ranking.…”

Towards automated correction of brain shift using deep deformable magnetic resonance imaging-intraoperative ultrasound (MRI-iUS) registration

“…Nevertheless, aligning preoperative MRI and iUS for brain shift correction is still a challenging problem due to the different characteristics of each modality and the type of information they provide. Consequently, only a few studies have applied deep learning on registering preoperative MRI and iUS for brain shift correction [9][10][11]. In this paper, a fast and robust deep learning-based method for automatic preoperative MRI and interventional US registration is presented to assist the neurosurgeons by correcting brain shift intraoperatively.…”

Deformable MRI-Ultrasound registration using correlation-based attribute matching for brain shift correction: Accuracy and generality in multi-site data

“…Pre-processing is needed in many MR-US registration studies. Some methods require skull stripping (Farnia, Ahmadian, Shabanian, Serej, & Alirezaie, 2014) (Farnia, Makkiabadi, Ahmadian, & Alirezaie, 2016) or tissue segmentation (Hong & Park, 2018) (Morin et al, 2017) (Palombi et al, 2018) (Reinertsen, Lindseth, Askeland, Iversen, & Unsgård, 2014) of MR images, or segmentation and removal of bright strips in skin surfaces (Wein, 2018) (Shams, Boucher, & Kadoury, 2018). These pre-processing steps are non-trivial, error-prone, and often require case-specific human intervention in tumorbearing MR images (Drobny et al, 2018) (Farnia et al, 2014) (Farnia et al, 2016).…”

“…A third contribution of our work is thorough evaluation. Most existing methods have been evaluated with only single-site data (Farnia et al, 2014) (Farnia et al, 2016) (Lindseth et al, 2003) (Farnia, Ahmadian, Shabanian, Serej, & Alirezaie, 2015), or up to two datasets but resulted in inconsistent levels of accuracy (Hong et al, 2018) (Shams, Boucher, & Kadoury, 2018) (Zhong et al, 2018), or used dataset-specific parameters (Wein, 2018) (Heinrich, 2018). We used three datasets, so far the most comprehensive multi-site data, to evaluate MR-US registration.…”