- Abstract: Intra-operative ultrasound is an increasingly important imaging modality in neuro- surgery. However, manual interaction with imaging data during the procedures, for example to select landmarks or perform segmentation, is difficult and can be time consuming. Yet, as registration to other imaging modalities is required in most cases, some annotation is necessary. We propose a segmentation method based on DeepVNet and specifically evaluate the integration of pre-training with simulated ultrasound sweeps to improve automatic segmentation and enable a fully automatic initialization of registration. Trained on coarse and incomplete semi-automatic annotations, our approach is able to capture the desired superficial structures such as sulci , the cerebellar tentorium , and the falx cerebri. We perform a five-fold cross-validation on the publicly available RESECT dataset. Trained on the dataset alone, we report a Dice and Jaccard coefficient of 0.45 ± 0.09 and 0.30 ± 0.07 respectively, as well as an average distance of 0.78 ± 0.36 mm. With the suggested pre-training, we computed a Dice and Jaccard coefficient of 0.47 ± 0.10 and 0.31 ± 0.08 , and an average distance of 0.71 ± 0.38 mm. The qualitative evaluation suggest that with pre-training the network can learn to generalize better and provide refined and more complete segmentations in comparison to incomplete annotations.
- Keywords: segmentation, intra-operative ultrasound, ultrasound, brain, 3D
- Author Affiliation: Technische Universität München, Technical University Munich, Johns Hopkins University